Canadian Centre for Cyber Security News

<article data-history-node-id="681" about="/en/guidance/how-use-online-banking-securely-itsap00080" class="cccs-basic-page full clearfix"> <div class="content"> <div class="layout layout–onecol"> <div class="layout__region layout__region–content"> <div data-block-plugin-id="extra_field_block:node:cccs_basic_page:links" class="block block-layout-builder block-extra-field-blocknodecccs-basic-pagelinks clearfix"> </div> <div data-block-plugin-id="field_block:node:cccs_basic_page:body" class="block block-layout-builder block-field-blocknodecccs-basic-pagebody clearfix"> <div class="field field–name-body field–type-text-with-summary field–label-hidden field–item"><div class="row"><!–Info across the top under the image–> <div class="col-md-4 pull-left hidden-xs hidden-sm"> <p class="text-left"><strong>October 2026</strong></p> </div> <div class="col-md-4 hidden-xs hidden-sm"> <p class="text-center"><strong>Awareness series</strong></p> </div> <div class="col-md-4 pull-right hidden-xs hidden-sm"> <p class="text-right"><strong>ITSAP.00.080</strong></p> </div> <!–MOBILE STARTS HERE–> <div class="hidden-lg hidden-md text-center"> <p><strong>October 2026 | Awareness series</strong></p> </div> <div class="col-md-12 mrgn-tp-lg"><!– <div class="mrgn-bttm-md well well-sm col-md-4 col-sm-12 col-12-xs pull-right mrgn-lft-md"> <p class="mrgn-tp-sm"><strong>Alternate format</strong>: <a href="/sites/default/files/itsap00080-r2-e.pdf">How to use online banking securely&nbsp;-&nbsp;ITSAP.00.080 (PDF,&nbsp;810&nbsp;KB)</a></p> </div>–> <p>Online banking allows for real-time access to your financial information. You can transfer money, pay bills and manage your accounts through a website or mobile app at any time. While it makes banking easier, it can also create opportunities for fraud and unauthorized access. Threat actors may try to steal your login information, redirect payments or trick you into approving transactions. As technology evolves, including the use of artificial intelligence (AI) in banking and online scams, it is important to understand how these risks can affect you.</p> <p>Being aware of how online banking works and the risks involved can help you use it securely. This publication explains common risks and steps you can take to stay safe when banking online.</p> <section><h2>On this page</h2> <ul><li><a href="#1">Common risks in online banking</a></li> <li><a href="#2">How to protect your banking information</a></li> <li><a href="#3">Steps to take if fraud occurs</a></li> <li><a href="#4">Learn more</a></li> </ul></section><section><h2 id="1">Common risks in online banking</h2> <p>Online banking is generally safe, but threat actors may still try to access your account in various ways. They may use a combination of several methods, which can lead to unauthorized bank account access or activity.</p> <h3>Phishing and social engineering</h3> <p>Phishing and other social engineering attacks try to manipulate you into sharing sensitive information or approving transactions. A threat actor may contact you by text message, email, phone call or through social media while pretending to be a trusted source, such as your bank.</p> <p>Common examples include:</p> <ul><li>messages claiming there is urgent or unusual activity on your account</li> <li>requests to confirm login details, one-time codes or personal information</li> <li>links or QR codes that lead to fake websites designed to look like your bank’s website</li> <li>messages that appear to come from real phone numbers or email addresses</li> <li>repeated sign-in approval requests designed to pressure you into accepting a login attempt you did not initiate</li> </ul><h3>Malware</h3> <p>Malware is software designed to access or control a device. If malware is installed on your phone, tablet or computer, a threat actor may be able to steal login information or redirect payments.</p> <p>This can happen when:</p> <ul><li>phishing messages include links or attachments that install malware</li> <li>fake or modified banking and money-transfer apps are downloaded from unofficial sources</li> <li>compromised websites or ads prompt you to install software or browser extensions</li> </ul><h3>Third-party apps and services</h3> <p>Online banking accounts can be linked to third-party apps or services, such as budgeting tools and payment services. They can also connect to retail store apps and digital wallets. These connections can be convenient but may introduce risks if access is not properly managed.</p> <p>This can happen when:</p> <ul><li>a connected app or service is compromised, exposing your personal and financial information that could then be used for scams</li> <li>your access remains active even after you are no longer using the app or service</li> <li>the permissions you set allow a third party to view your account information or initiate actions you did not expect</li> <li>you mistake messages from a connected service for legitimate communication from your bank</li> </ul><h3><abbr title="subscriber identity module">SIM</abbr> swapping</h3> <p>Subscriber identity module (SIM) swapping is a type of fraud in which a threat actor takes control of your mobile number by transferring it to a <abbr title="subscriber identity module">SIM</abbr> card they control. The threat actor can then receive your calls and text messages, including one-time passcodes used for account access and verification. Once they control your number, they may be able to reset passwords, access your online banking and make unauthorized transactions.</p> <p>This can happen when a threat actor:</p> <ul><li>gathers personal information through phishing, data breaches or social media</li> <li>impersonates you and convinces your mobile provider to transfer your number to a new <abbr title="subscriber identity module">SIM</abbr> card</li> </ul><p>Someone may have swapped your <abbr title="subscriber identity module">SIM</abbr> card if:</p> <ul><li>your device suddenly loses mobile service, preventing you from receiving calls or messages</li> <li>you stop receiving expected text messages, including one-time passcodes or account recovery messages</li> <li>you are locked out of your banking, email or other accounts while fraudulent activity occurs</li> </ul><section class="panel panel-default"><header class="panel-heading"><h3 class="panel-title">How <abbr title="artificial intelligence">AI</abbr> is changing online banking scams</h3> </header><div class="panel-body"> <p><abbr title="artificial intelligence">AI</abbr> is helping banks improve security by detecting unusual activity and preventing fraud. At the same time, threat actors are using AI to make scams more convincing and harder to detect.</p> <p>Threat actors can use <abbr title="artificial intelligence">AI</abbr> to:</p> <ul><li>create realistic emails, text messages or chat conversations that appear to come from your bank</li> <li>generate fake websites or apps that look like legitimate services</li> <li>impersonate voices or trusted contacts to request money or sensitive information</li> <li>tailor scams using publicly available information about you</li> <li>automate attacks and target more people quickly</li> </ul><p>Understanding how these <abbr title="artificial intelligence">AI</abbr> tactics are used in scams can help you spot warning signs and make safer decisions when banking online.</p> </div> </section></section><section><h2 id="2">How to protect your banking information</h2> <p>You can take actions to reduce the risks associated with online banking. Good security habits help protect your account, personal information and money, especially when you bank on mobile devices and use real-time payments.</p> <h3>Understand how your bank protects you</h3> <p>Recognize normal activity and spot anything unusual by learning what your bank offers.</p> <ul><li>Learn about your bank’s security features, alerts and notification options</li> <li>Understand how your bank will contact you about suspicious activity</li> <li>Review your account settings and online banking terms so you know what you’re responsible for and what your bank will cover if an incident occurs</li> </ul><h3>Secure your devices and access</h3> <p>Protecting the devices you use for banking is one of the best ways to prevent unauthorized access.</p> <ul><li>Lock your devices with a personal identification number (PIN), passphrase or biometric feature</li> <li>Use a strong, unique password or passphrase for your bank account</li> <li>Use multi-factor authentication to provide an extra layer of security if your passphrase is compromised</li> </ul><h3>Keep your systems and apps up to date</h3> <p>Regular updates help fix security issues that threat actors often exploit.</p> <ul><li>Use security software where appropriate and enable built-in device protections</li> <li>Update and patch your device and software to fix security vulnerabilities</li> <li>Download banking apps only from official app stores or your financial institution’s website</li> <li>Use trusted Internet connections like your home network, mobile data and trusted VPNs, as they provide more protection than public Wi-Fi</li> </ul><h3>Monitor your account and activity</h3> <p>Checking your account regularly helps you spot and respond to issues quickly.</p> <ul><li>Regularly review transactions, alerts and account changes</li> <li>Turn on notifications for account logins, payments and updates</li> <li>Pay attention to unexpected alerts or blocked transactions, as these can be early signs of fraud</li> </ul><h3>Recognize and avoid common risks</h3> <p>Being aware of common tactics used by threat actors can help you avoid mistakes.</p> <ul><li>Never share your login details, one-time codes or security answers with anyone, including third-party apps and services</li> <li>Do not use personal details such as your birthdate or pet’s name in passphrases or security questions</li> <li>Do not store banking information in unsecured places like notes or plain text files. If you do need to store or save banking information, make sure you protect it with encryption</li> <li>Do not open links, attachments, QR codes or pop-ups in a message unless you have confirmed that the message is legitimate</li> <li>Do not access your bank account from public or shared computers or networks</li> <li>Only approve sign-in or payment requests that you initiated</li> </ul><div class="well"> <p>If you receive a message or call claiming to be from your bank and you are unsure whether it is legitimate:</p> <ul><li>do not reply to the message or open any links</li> <li>contact your bank directly using the phone number on the back of your bank card or through your official banking app or website</li> <li>check your account activity and alerts for anything unexpected</li> </ul><p>When in doubt, always use a trusted way to contact your bank rather than responding to the message you received.</p> </div> </section><section><h2 id="3">Steps to take if fraud occurs</h2> <p>If you believe your online banking account has been compromised, or you think you have mistakenly approved or sent a payment, <strong>act quickly</strong> to reduce further harm.</p> <h3>Step 1: Contact your bank</h3> <ul><li>Contact your bank right away using the phone number on the back of your bank card or through an official banking app or website</li> <li>Ask the bank to secure your account, which may include locking access, stopping or reversing transactions and resetting your login details</li> </ul><h3>Step 2: Review your accounts and information</h3> <ul><li>Check your recent transactions for anything you do not recognize</li> <li>Review your contact information, security settings and linked devices or services to make sure no changes were made without your approval</li> <li>Watch for follow-up messages or calls, as additional scams may occur after a fraud attempt</li> </ul><h3>Step 3: Report the incident and monitor your accounts</h3> <ul><li>Report fraud or attempted fraud via the <a href="https://antifraudcentre-centreantifraude.ca/index-eng.htm">Canadian Anti-Fraud Centre website</a> or by calling <a href="tel:+1-888-495-8501">1-888-495-8501</a></li> <li>File a report with local law enforcement if your accounts were accessed without your permission</li> <li>Monitor your accounts and alerts closely in the days and weeks following the incident</li> <li>Report any new or suspicious activity to your bank immediately</li> </ul></section><section><h2 id="4">Learn more</h2> <ul><li><a href="/en/guidance/dont-take-bait-recognize-and-avoid-phishing-attacks">Don’t take the bait: Recognize and avoid phishing attacks (ITSAP.00.101)</a></li> <li><a href="/en/guidance/protect-your-organization-malware-itsap00057">Protect your organization from malware (ITSAP.00.057)</a></li> <li><a href="/en/guidance/best-practices-passphrases-and-passwords-itsap30032">Best practices for passphrases and passwords (ITSAP.30.032)</a></li> <li><a href="/en/guidance/secure-your-accounts-and-devices-multi-factor-authentication-itsap30030">Secure your accounts and devices with multi-factor authentication (ITSAP.30.030)</a></li> <li><a href="/en/guidance/preventative-security-tools-itsap00058">Preventative security tools (ITSAP.00.058)</a></li> <li><a href="/en/guidance/how-updates-secure-your-device-itsap10096">How updates secure your device (ITSAP.10.096)</a></li> <li><a href="https://www.canada.ca/en/financial-consumer-agency/services/banking/artificial-intelligence.html">Artificial intelligence in banking – Canada.ca</a></li> <li><a href="/en/guidance/security-considerations-sims">Security considerations for <abbr title="subscriber identity modules">SIMs</abbr> (ITSAP.10.021)</a></li> </ul></section></div> </div> </div> </div> </div> </div> </div> </article>

<article data-history-node-id="8364" about="/en/news-events/joint-guidance-prc-government-linked-cyber-threat-actors-combining-automated-hands-hacking-tools-steal-sensitive-data" class="cccs-basic-page full clearfix"> <div class="content"> <div class="layout layout–onecol"> <div class="layout__region layout__region–content"> <div data-block-plugin-id="extra_field_block:node:cccs_basic_page:links" class="block block-layout-builder block-extra-field-blocknodecccs-basic-pagelinks clearfix"> </div> <div data-block-plugin-id="field_block:node:cccs_basic_page:body" class="block block-layout-builder block-field-blocknodecccs-basic-pagebody clearfix"> <div class="field field–name-body field–type-text-with-summary field–label-hidden field–item"><p>The Canadian Centre for Cyber Security (Cyber Centre) has joined the United States’ Federal Bureau of Investigation (FBI) and the following international partners in releasing cyber security guidance on the Peoples Republic of China (PRC) government-linked threat actors combining automated and hands-on hacking tools to steal sensitive information:</p> <ul><li>Cybersecurity and Infrastructure Security Agency (CISA)</li> <li>National Security Agency (NSA)</li> <li>United Kingdom’s National Cyber Security Centre (NCSC-UK)</li> <li>New Zealand’s National Cyber Security Centre (NCSC-NZ)</li> <li>Australian Signals Directorate (ASD)</li> <li>Japan’s National Police Agency (NPA)</li> <li>Japan’s National Cybersecurity Office (NCO)</li> <li>Spain’s Centro Nacional de Inteligencia (CNI)</li> </ul><p>This joint guidance details how <abbr title="Peoples Republic of China">PRC</abbr>-linked threat actors exploit networks across multiple sectors. It provides organizations an analysis of tactics, techniques and procedures (TTPs) and indicators of compromise (IOCs) from the threats posed by these actors.</p> <p>Network defenders can apply detection and mitigation strategies that reduce the risks of threat actors compromising critical data. Organizations can take the following immediate actions to protect against <abbr title="Peoples Republic of China">PRC</abbr>-linked cyber threat activity:</p> <ul><li>Disable unused services and ports, such as automatic configuration, remote access or file sharing protocols</li> <li>Sanitize user input in web applications to prevent possible cross-site scripting (XSS) payload injection</li> <li>Implement identity, credential and access management (ICAM) policies and require multi-factor authentication (MFA) for services, where possible</li> </ul><p>Consult the full joint guidance: <a href="https://www.ic3.gov/CSA/2026/261008.pdf">Chinese Government-linked Cyber Threat Actors Combine Automated and Hands-on Hacking Tools to Steal Sensitive Data (PDF)</a></p> </div> </div> </div> </div> </div> </article>

This technical article is the second in a series on the work of the Frontier AI Lab. This second article focuses on malware reverse engineering, the process of deconstructing malicious software without its original source code to help cyber defenders understand its functionality and purpose.

<article data-history-node-id="631" about="/en/guidance/shopping-online-safely-itsap00071" class="cccs-basic-page full clearfix"> <div class="content"> <div class="layout layout–onecol"> <div class="layout__region layout__region–content"> <div data-block-plugin-id="extra_field_block:node:cccs_basic_page:links" class="block block-layout-builder block-extra-field-blocknodecccs-basic-pagelinks clearfix"> </div> <div data-block-plugin-id="field_block:node:cccs_basic_page:body" class="block block-layout-builder block-field-blocknodecccs-basic-pagebody clearfix"> <div class="field field–name-body field–type-text-with-summary field–label-hidden field–item"><section><div class="row"><!–Info across the top under the image–> <div class="col-md-4 pull-left"> <p class="text-left"><strong>October 2026</strong></p> </div> <div class="col-md-4"> <p class="text-center"><strong>Awareness series</strong></p> </div> <div class="col-md-4 pull-right"> <p class="text-right"><strong>ITSAP.00.071</strong></p> </div> <!–ENGLISH Intro paragraph plus pdf download–> <div class="col-md-12 mrgn-tp-lg"><!– <div class="mrgn-bttm-md well well-sm col-md-4 pull-right mrgn-lft-md"> <p class="mrgn-tp-sm"><strong>Alternate format</strong>: <a href="/sites/default/files/itsap00071-how-shop-online-safely-e.pdf">How to shop online safely&nbsp;- ITSAP.00.071 (PDF,&nbsp;832&nbsp;KB)</a></p> </div> –> <p>E-commerce is evolving and so have the threats that target shoppers and the information they rely on.</p> <p>Threat actors use increasingly convincing techniques to steal personal and financial information, redirect payments and compromise online accounts. These activities can put both your personal information and your organization’s assets at risk, including email accounts, shipping details, phone numbers and credit card information.</p> <p>Whether you are shopping for yourself or making purchases for your organization, taking a few simple precautions can help you stay safe online.</p> </div> </div> </section><details><summary><h2 class="h3">On this page</h2> </summary><ul><li><a href="#Common-threats">Understand common online shopping threats</a></li> <li><a href="#Signs-scam">Recognize the warning signs of a scam</a></li> <li><a href="#Protect-yourself">Protect yourself when shopping online</a></li> <li><a href="#Steps-take">Steps to take after an online shopping incident</a></li> <li><a href="#Learn">Learn more</a></li> </ul></details><section><h2 id="Common-threats">Understand common online shopping threats</h2> <p>Online shopping could put you and your organization at risk of identity theft, account compromise and financial loss. Threat actors now use a wider range of techniques to make fraudulent websites, ads and messages appear legitimate and trustworthy. Some of these techniques may be automated or enhanced using artificial intelligence (AI), making scams harder to detect. The following examples demonstrate common ways threat actors can gain access to user accounts and obtain sensitive information:</p> <ul><li>fake e-commerce sites that collect your information or payment details after you complete a purchase without delivering goods or services</li> <li>fraudulent payment processing sites, including fake checkout pages or third-party payment requests, that collect money for fake or unauthorized transactions</li> <li>websites that do not use encrypted connections (such as, HTTPS), increasing the risk of your information being intercepted, particularly when you use public or unsecured networks</li> <li>websites or online marketplaces that lack basic consumer protections and do not verify sellers, increasing the risk of fraud</li> <li>impersonation scams where criminals copy the names, logos, and designs of real retailers or delivery companies</li> <li>fraudulent shopping links or QR codes delivered through malicious shopping apps, social media ads, emails or text messages</li> <li>requests to move payments outside the official checkout process, such as paying by e-transfer, gift card or cryptocurrency, or to return money for an alleged overpayment</li> </ul><p>Professional-looking websites, padlock icons, encrypted connections and familiar brand names do not guarantee a site is legitimate. Many fraudulent shopping sites are designed to closely resemble trusted retailers and exploit the trust users have in familiar brands.</p> <div class="panel panel-default"> <header class="panel-heading"><h5 class="panel-title">How AI is affecting online shopping</h5> </header><div class="panel-body"> <p><abbr title="artificial intelligence">AI</abbr> is increasingly used to create shopping websites, generate advertisements, recommend products and automate customer interactions. While these technologies can improve convenience, they also introduce new risks for online shoppers. <abbr title="artificial intelligence">AI</abbr> can be used by both legitimate businesses and scammers, making it more difficult to distinguish between genuine and fraudulent content.</p> <p><strong>AI-enabled shopping scams</strong></p> <p>Scammers can use <abbr title="artificial intelligence">AI</abbr> to create convincing websites, advertisements, product descriptions, reviews and customer service interactions. <abbr title="artificial intelligence">AI</abbr>-generated content can appear professional and authentic, making scams more difficult to spot. In the past, poor grammar, spelling mistakes or low-quality website content were common warning signs. While these signs can still indicate fraud, they may be less reliable than before.</p> <p><strong>Using AI shopping assistants</strong></p> <p>Some online tools now act as autonomous or semi-autonomous shopping assistants, making recommendations or completing purchases on a buyer’s behalf. Relying too heavily on these tools may increase the risk of:</p> <ul><li>placing too much trust in <abbr title="artificial intelligence">AI</abbr> recommendation or decisions, even when they are based on incomplete, outdated or manipulated information</li> <li>automated decisions being made without sufficient verification, such as choosing unfamiliar sellers or responding to time-limited offers</li> <li>sharing personal, financial or browsing data with <abbr title="artificial intelligence">AI</abbr> tools that may store, reuse or expose this information if not properly secured</li> <li>being influenced by recommendations that prioritize sponsored, manipulated or artificially promoted products over legitimate options</li> </ul><p><abbr title="artificial intelligence">AI</abbr> tools cannot reliably determine whether a seller, website or offer is trustworthy or fraudulent in the same way a person can. You should remain actively involved in shopping decisions and verify sellers and payment details. Treat <abbr title="artificial intelligence">AI</abbr>-generated recommendations as guidance, not assurances of legitimacy.</p> </div> </div> </section><section><h2 id="Signs-scam">Recognize the warning signs of a scam</h2> <p>Fraudulent shopping websites, including those created with <abbr title="artificial intelligence">AI</abbr> tools, can appear legitimate and convincing. Use the following checks to help assess whether a website is safe to use.</p> <h3>Check how the website looks and behaves</h3> <ul><li>Look for poor design elements that make the site appear unprofessional</li> <li>Observe whether links, buttons and navigation features seem to work properly</li> <li>Check for clear contact information such as an address, email or phone number</li> <li>Review the privacy, return and refund policies and make sure they are clear</li> <li>Question requests for personal or payment information that is not needed for your purchase</li> </ul><h3>Be careful of unusual offers and suspicious messages</h3> <ul><li>Compare prices and be cautious of deals that seem too good to be true</li> <li>Review shipping costs, duties and extra charges for anything that seems abnormal, and be cautious of unexpected fees that significantly increase the total cost</li> <li>Notice pressure tactics that push you to act quickly, such as limited time or last-chance messages</li> <li>Be alert to unexpected follow-up messages or customer-support requests claiming issues with your payment, delivery, customs fees, refunds or order status that require urgent action</li> </ul></section><section><h2 id="Protect-yourself">Protect yourself when shopping online</h2> <p>The steps below can help protect you and your organization when shopping online. While these practices reduce risk, no single step can eliminate it completely.</p> <h3>Research the retailer</h3> <p>Before making a purchase:</p> <ul><li>search for independent reviews beyond the retailer’s website</li> <li>check how long the retailer has been operating and whether it has an established online presence</li> <li>look for complaints, scam reports or unresolved customer issues</li> <li>confirm that customer service channels respond to inquiries before making large purchases</li> </ul><p>When a product is advertised through a social media platform or marketplace:</p> <ul><li>confirm whether the seller or account is verified by the platform</li> <li>review feedback from multiple verified or confirmed buyers and look for consistent patterns across reviews</li> <li>use reputable online marketplaces with clear return and fraud-protection policies for third-party resellers</li> </ul><h3>Browse safely</h3> <p>When visiting shopping websites:</p> <ul><li>configure <a href="https://www.cira.ca/en/">CIRA Canadian Shield</a> on your home network or devices to help block known malicious websites</li> <li>use websites that support encrypted connections (HTTPS) to help protect your information</li> <li>follow browser security warnings and do not continue to websites flagged as unsafe</li> <li>do not rely on padlock icons or website appearance alone to determine if the site is legitimate</li> <li>be careful when shopping on mobile devices, where shortened URLs and limited screen space can make fraudulent sites harder to identify</li> </ul><h3>Protect your accounts</h3> <p>To reduce account and payment-related risks:</p> <ul><li>create a unique and strong passphrase for each online account</li> <li>avoid reusing passphrases for shopping, email and financial accounts</li> <li>use phishing-resistant multi-factor authentication (MFA) where available</li> <li>check your credit card statements regularly for unfamiliar charges</li> <li>use a safe form of payment from major financial institutions, such as credit cards, that offer fraud protection or chargeback options</li> <li>use virtual credit card numbers or accounts with spending limits when shopping with unfamiliar retailers</li> </ul><h3>Take additional precautions</h3> <p>To further reduce your exposure, you should:</p> <ul><li>limit the amount of personal information you provide to what is strictly necessary</li> <li>use guest checkout when available to reduce the likelihood that the retailer will store (or share) your personal information. <ul><li>if you create an account, protect it with a strong passphrase and <abbr title="multi-factor authentication">MFA</abbr></li> </ul></li> <li>regularly back up your devices and update them with the latest security patches <ul><li>use secure, password-protected Wi-Fi networks and avoid making purchases on public Wi-Fi</li> </ul></li> <li>be cautious of promotional emails, texts or messages offering special deals</li> <li>go directly to retailers’ official websites instead of clicking links</li> </ul></section><section><h2 id="Steps-take">Steps to take after an online shopping incident</h2> <p>Threats and scams can be difficult to spot. You may not even know that you’ve been targeted until after your money has been stolen or information has already been shared. Even if you’re not sure if you’ve been scammed, it is important to act quickly. If you believe you are the victim of a scam or a potential compromise, you should take the following actions to report and mitigate the incident.</p> <h3>Step 1: Notify your organization</h3> <p>If the purchase was made on behalf of your organization or involved organizational accounts, devices or payment methods, report the incident to your security department, technical support or senior management. Early reporting helps reduce further risk and supports organizational response and recovery.</p> <h3>Step 2: Contact your credit card company</h3> <p>Report the suspicious transaction immediately. Ask your bank or credit card company to cancel or reverse the charge and clarify whether account monitoring or a replacement card is needed.</p> <h3>Step 3: Reset affected accounts</h3> <p>Change passwords for any affected shopping, email and social media accounts. Use strong, unique passphrases and turn on phishing-resistant <abbr title="multi-factor authentication">MFA</abbr> where available. If you used the same password on other sites or accounts, change it on those accounts too and use unique passwords for each of your accounts going forward.</p> <h4>Step 4: Report the incident</h4> <p>Report fraud and cybercrime to the <a href="https://reportcyberandfraud.canada.ca/">Canadian Anti-Fraud Centre</a> or by calling <a href="tel:+1-888-495-8501">1-888-495-8501</a>. You should also report attempted scams even if you didn’t lose any money, as this information helps protect others.</p> </section><section><h2 id="Learn">Learn more</h2> <ul><li><a data-entity-substitution="canonical" data-entity-type="node" data-entity-uuid="300c9e0e-f1b5-4344-97b3-e98cc202ca32" href="/en/guidance/using-encryption-keep-your-sensitive-data-secure-itsap40016">Using encryption to keep your sensitive data secure (ITSAP.40.016)</a></li> <li><a data-entity-substitution="canonical" data-entity-type="node" data-entity-uuid="32369248-5885-44f2-9af9-d4a80bb9c8b6" href="/en/guidance/best-practices-passphrases-and-passwords-itsap30032">Best practices for passphrases and passwords (ITSAP.30.032)</a></li> <li><a data-entity-substitution="canonical" data-entity-type="node" data-entity-uuid="df4d897c-c726-4e48-8901-408ba2bdf6d3" href="/en/guidance/secure-your-accounts-and-devices-multi-factor-authentication-itsap30030">Secure your accounts and devices with multi-factor authentication (ITSAP.30.030)</a></li> <li><a data-entity-substitution="canonical" data-entity-type="node" data-entity-uuid="639ede31-0d64-45cb-9688-664b3ee445cb" href="/en/guidance/how-updates-secure-your-device-itsap10096">How updates secure your device (ITSAP.10.096)</a></li> <li><a data-entity-substitution="canonical" data-entity-type="node" data-entity-uuid="481aa714-81de-40f1-a235-4b363a33b5ed" href="/en/guidance/virtual-private-networks-itsap80101">Virtual private networks (ITSAP.80.101)</a></li> <li><a data-entity-substitution="canonical" data-entity-type="node" data-entity-uuid="719fcafc-bc2d-4270-abb3-b8933711ceb4" href="/en/blogs/how-spot-unsafe-online-store">How to spot an unsafe online store</a></li> </ul></section></div> </div> </div> </div> </div> </article>

This publication provides insight into the threat landscape and outlines corresponding mitigation strategies to safeguard the integrity and reliability of cable infrastructure.

<article data-history-node-id="8302" about="/en/news-events/statement-regarding-reported-activity-targeting-government-canada-websites" class="cccs-basic-page full clearfix"> <div class="content"> <div class="layout layout–onecol"> <div class="layout__region layout__region–content"> <div data-block-plugin-id="extra_field_block:node:cccs_basic_page:links" class="block block-layout-builder block-extra-field-blocknodecccs-basic-pagelinks clearfix"> </div> <div data-block-plugin-id="field_block:node:cccs_basic_page:body" class="block block-layout-builder block-field-blocknodecccs-basic-pagebody clearfix"> <div class="field field–name-body field–type-text-with-summary field–label-hidden field–item"><p><strong>From:</strong> <a href="https://www.cse-cst.gc.ca/en"><strong>Communications Security Establishment Canada</strong></a></p> <h2>Statement</h2> <p><strong>September 29, 2026</strong> – Ottawa, Ontario</p> <p>We are aware of reports identifying suspicious activity, including suspected AI agent activity, targeting publicly accessible websites, such as the Government of Canada. There is no indication that government systems have been compromised at this time.</p> <p>Public-facing government websites routinely receive automated and potentially malicious requests. Such activity is an ongoing feature of the online environment and does not, on its own, indicate a successful cyber incident. While such activity is common, the Government of Canada takes reports of potential cyber security incidents and suspicious online activity seriously. The Canadian Centre for Cyber Security (Cyber Centre) is working closely with government partners to assess the information referenced in the reports.</p> <p>The Cyber Centre employs robust, layered cyber security measures to help protect Government of Canada networks, systems and information from evolving cyber threats. Our world-class cyber defence tools are designed to detect and block malicious activity, regardless of its source.</p> <p>More broadly, the Cyber Centre works closely with partners in Canada and internationally, including all levels of government, Indigenous communities, industry, academia and developers of advanced AI models, to better understand the opportunities, challenges and cyber security risks associated with emerging technologies such as artificial intelligence. Through this work, the Cyber Centre shares advice, guidance, threat intelligence and cyber defence services to help strengthen the security and resilience of Government of Canada systems, critical infrastructure and Canada’s broader digital ecosystem.</p> <h3>Contacts</h3> <p>For more information, please contact (media only):</p> <h4>CSE Media Relations</h4> <p><a href="mailto:media@cse-cst.gc.ca">media@cse-cst.gc.ca</a></p> <h4>Stay connected</h4> <p>Follow <a href="https://x.com/CyberCentre_ca">Cyber Centre</a> and <a href="https://x.com/cse_cst">CSE</a> on X (formerly known as Twitter) and on <a href="https://www.linkedin.com/company/cse_cst/">LinkedIn</a>.</p> </div> </div> </div> </div> </div> </article>

The time to act is now: Strengthening critical infrastructure cyber readiness for a resilient Canada.

As cyber threats become more complex and threat actors more sophisticated, many security operations centres are being set up.

This guidance provides strategies to prevent, detect and mitigate prevalent Active Directory compromises

<article data-history-node-id="8188" about="/en/news-events/cyber-threat-actors-use-artificial-intelligence-active-global-campaign-disrupt-internet-exposed-programmable-logic-controllers" class="cccs-basic-page full clearfix"> <div class="content"> <div class="layout layout–onecol"> <div class="layout__region layout__region–content"> <div data-block-plugin-id="extra_field_block:node:cccs_basic_page:links" class="block block-layout-builder block-extra-field-blocknodecccs-basic-pagelinks clearfix"> </div> <div data-block-plugin-id="field_block:node:cccs_basic_page:body" class="block block-layout-builder block-field-blocknodecccs-basic-pagebody clearfix"> <div class="field field–name-body field–type-text-with-summary field–label-hidden field–item"><p>The Canadian Centre for Cyber Security (Cyber Centre) is warning Canadian organizations of an active global campaign in which cyber threat actors are targeting and attempting to disrupt programmable logic controllers (PLCs) that are exposed to the internet. The Cyber Centre and its partners have previously warned that cyber threat actors are targeting internet-accessible industrial control systems and operational technologies (<abbr title="Industrial Control Systems">ICS</abbr>/<abbr title="Operational Technology">OT</abbr>) to disrupt water facilities and other critical services in Canada and abroad. This cybersecurity advisory describes how threat actors are using artificial intelligence (AI) to find and attack these devices.</p> <h2>The threat to Canadian organizations</h2> <p>The Cyber Centre continues to observe cyber threat actors targeting and attempting to disrupt internet-exposed <abbr title="Industrial Control Systems">ICS</abbr>/<abbr title="Operational Technology">OT</abbr> systems in Canada. In October 2025, the Cyber Centre observed several incidents affecting <abbr title="Industrial Control Systems">ICS</abbr>/<abbr title="Operational Technology">OT</abbr> systems in Canada. In one incident, a threat actor manipulated water pressure valve controls at a municipal water system, causing minor impacts to water service delivery.</p> <p>In July 2026, the United States’ Cybersecurity and Infrastructure Security Agency (CISA) reported a significant increase in attacks against <abbr title="Programmable Logic Controllers">PLCs</abbr> at water and wastewater utilities. Threat actors changed passwords to lock operators out of their controllers and took devices offline by changing their network addresses. This activity resulted in boil water advisories and sustained manual operations at affected utilities.</p> <p>Threat actors are testing and refining their exploitation techniques against specific PLC models. They are focusing their activity on specific sectors and facilities. Our partners assess that this activity is likely persistent reconnaissance intended to develop capabilities and prepare to cause operational effects against critical infrastructure. The most targeted sectors include: critical manufacturing, energy, water and wastewater, chemical, food and agriculture, and commercial facilities. Threat actors are targeting organizations of every size.</p> <p>Canada has not experienced a high-impact <abbr title="Industrial Control Systems">ICS</abbr>/<abbr title="Operational Technology">OT</abbr> incident so far. However, we assess that the threat to <abbr title="Industrial Control Systems">ICS</abbr>/<abbr title="Operational Technology">OT</abbr> owners and operators is increasing and that future incidents will likely be more severe.</p> <p>To monitor and mitigate this threat, we encourage Canadian organizations to consult the guidance linked below on reducing internet exposure, securing remote access, and protecting <abbr title="Industrial Control Systems">ICS</abbr>/<abbr title="Operational Technology">OT</abbr> systems.</p> <h2>The threat to internet-exposed <abbr title="Programmable Logic Controllers">PLCs</abbr></h2> <p><abbr title="Programmable Logic Controllers">PLCs</abbr> are small industrial computers that control physical processes, such as pumps, valves, pressure, and temperature. They are used in facilities across every critical infrastructure sector. A threat actor with access to a PLC can disrupt industrial processes, damage equipment, endanger worker and public safety. The impacts can spread across interconnected systems. While <abbr title="Programmable Logic Controllers">PLCs</abbr> are a frequent target because they directly control physical processes, other internet-exposed <abbr title="Industrial Control Systems">ICS</abbr>/<abbr title="Operational Technology">OT</abbr> assets, including human-machine interfaces (HMIs), are also at risk.</p> <p>Cyber threat actors use internet scanning services to find exposed <abbr title="Programmable Logic Controllers">PLCs</abbr> and take advantage of weak, default, or absent passwords to access them. After gaining access, threat actors have changed device settings, disabled alarms, and locked operators out of their systems, forcing facilities into manual operations while service is restored.</p> <p>Many <abbr title="Programmable Logic Controllers">PLCs</abbr> are exposed to the internet unintentionally and without the knowledge of the organizations that own them. These systems can be readily identified using public internet scanning services. A control system may be isolated from an organization’s enterprise information technology (IT) network but still be accessible from the internet. A cellular modem or a vendor remote access connection installed during commissioning or maintenance may provide access. These connections are often undocumented, retained after they are no longer required, and missed by routine attack surface scans. Organizations should verify that their systems cannot be accessed from the internet.</p> <p>Organizations should also identify and secure all internet-accessible <abbr title="Industrial Control Systems">ICS</abbr>/<abbr title="Operational Technology">OT</abbr> assets, not only <abbr title="Programmable Logic Controllers">PLCs</abbr>, and ensure remote access is properly controlled and monitored.</p> <h2>The risk from third parties and system integrators</h2> <p>The risk of unknown exposure is greatest for organizations that rely on system integrators or third-party service providers who may have remote access to their <abbr title="Programmable Logic Controllers">PLCs</abbr> for maintenance, monitoring, or support activities. Even organizations with mature cyber security programs should validate their external connections with their integrators and service providers and scan the internet address ranges they own to identify exposed systems, especially connections set up by third parties that are no longer actively managed or documented.</p> <p>Unclear division of roles and responsibilities between organizations and their service providers often creates gaps that leave critical systems unprotected. We encourage organizations to work closely with their integrators and service providers to ensure that systems are implemented securely and maintained throughout their lifecycle, and to consider including maintenance, security responsibilities, and liability in service agreements, in alignment with security best practices.</p> <h2>Cyber threat actors use artificial intelligence to attack <abbr title="Programmable Logic Controllers">PLCs</abbr></h2> <p>As we noted in our cyber threat bulletin on non-state activity targeting Canadian operational technology, AI tools capable of supporting cyber threat activity are rapidly improving and becoming increasingly available. These tools allow even low-sophistication threat actors to conduct more complex and disruptive attacks against <abbr title="Industrial Control Systems">ICS</abbr>/<abbr title="Operational Technology">OT</abbr> systems. We assess that as these tools become more widely available, opportunistic attacks against internet-exposed systems will likely also become more frequent.</p> <p>As part of this campaign, threat actors are using AI to generate exploitation scripts that they disguise as legitimate monitoring software, dramatically reducing the expertise and time needed to develop working attack tools. It is very likely that the actors continue to operate despite public reporting outlining their activities. Our partners issued their most recent warnings in August 2026.</p> <h2>Useful resources</h2> <p>Refer to the following resources for more information and useful advice and guidance.</p> <h3>Reports and advisories</h3> <ul><li>Canada’s threat assessments <ul><li><a href="/en/guidance/national-cyber-threat-assessment-2025-2026">National Cyber Threat Assessment 2025-2026</a></li> </ul></li> <li>Advisories and partner publications <ul><li><a href="/en/guidance/cyber-threat-bulletin-non-state-activity-targeting-canadian-operational-technology">Cyber threat bulletin: Non-state activity targeting Canadian operational technology</a></li> <li><a href="/en/alerts-advisories/al25-016-internet-accessible-industrial-control-systems-ics-abused-hacktivists">Alert AL25-016 Internet-accessible industrial control systems (ICS) abused by hacktivists</a></li> <li><a href="https://www.cisa.gov/news-events/cybersecurity-advisories/aa26-231a">Defending Against an Active Threat to Siemens S7 Series <abbr title="Programmable Logic Controllers">PLCs</abbr></a></li> <li><a href="https://www.cisa.gov/news-events/alerts/2026/07/30/cisa-urges-water-and-wastewater-systems-sector-protect-ot-against-activity-targeting-plcs"><abbr title="Cybersecurity and Infrastructure Security Agency">CISA</abbr> Urges Water and Wastewater Systems Sector to Protect OT Against Activity Targeting <abbr title="Programmable Logic Controllers">PLCs</abbr></a></li> <li><a href="https://www.ncsc.gov.uk/news/disruptive-cyber-activity-highlights-risk-from-internet-exposed-systems-and-edge-devices">Disruptive cyber activity highlights risk from internet-exposed systems and edge devices</a></li> <li><a href="https://www.cisa.gov/news-events/cybersecurity-advisories/aa25-343a">Pro-Russia Hacktivists Conduct Opportunistic Attacks Against US and Global Critical Infrastructure</a></li> <li><a href="https://www.ncsc.gov.uk/collection/operational-technology/definitive-architecture-view">Creating and maintaining a definitive view of your <abbr title="Operational Technology">OT</abbr> architecture</a></li> <li><a href="https://www.ncsc.gov.uk/collection/operational-technology/secure-connectivity">Secure connectivity principles for operational technology (OT)</a></li> <li><a href="https://www.cisa.gov/resources-tools/resources/primary-mitigations-reduce-cyber-threats-operational-technology">Primary Mitigations to Reduce Cyber Threats to Operational Technology</a></li> <li><a href="https://www.cisa.gov/resources-tools/resources/exposure-reduction">Internet Exposure Reduction Guidance</a></li> </ul></li> </ul><h3>Advice and guidance</h3> <ul><li><a href="/en/cyber-security-readiness/cyber-security-readiness-goals-securing-our-most-critical-systems">Cyber Security Readiness Goals: Securing Our Most Critical Systems</a></li> <li><a href="/en/cyber-security-readiness/cross-sector-cyber-security-readiness-goals-toolkit">Cross-Sector Cyber Security Readiness Goals Toolkit</a></li> <li><a href="/en/cyber-security-readiness/critical-infrastructure-resilience-escalated-threat-navigation-initiative">Critical infrastructure resilience and escalated threat navigation initiative</a></li> <li><a href="/en/guidance/security-considerations-critical-infrastructure-itsap10100">Security considerations for critical infrastructure (ITSAP.10.100)</a></li> <li><a href="/en/guidance/protect-your-operational-technology-itsap00051">Protect your operational technology (ITSAP.00.051)</a></li> <li><a href="/en/guidance/security-considerations-industrial-control-systems-itsap00050">Security considerations for industrial control systems (ITSAP.00.050)</a></li> <li><a href="/en/guidance/frontier-artificial-intelligence-itsap10050">Frontier artificial intelligence (ITSAP.10.050)</a></li> </ul></div> </div> </div> </div> </div> </article>

<article data-history-node-id="8109" about="/en/guidance/cyber-security-privacy-risk-management/organizational-cyber-security-privacy-risk-management-activities-itsp10036" class="cccs-basic-page full clearfix"> <div class="content"> <div class="layout layout–onecol"> <div class="layout__region layout__region–content"> <div data-block-plugin-id="extra_field_block:node:cccs_basic_page:links" class="block block-layout-builder block-extra-field-blocknodecccs-basic-pagelinks clearfix"> </div> <div data-block-plugin-id="field_block:node:cccs_basic_page:body" class="block block-layout-builder block-field-blocknodecccs-basic-pagebody clearfix"> <div class="field field–name-body field–type-text-with-summary field–label-hidden field–item"><div class="row"> <div class="col-md-4 pull-left hidden-xs hidden-sm"> <p class="text-left"><strong>September 2026</strong></p> </div> <div class="col-md-4 hidden-xs hidden-sm"> <p class="text-center"><strong>Practitioner series</strong></p> </div> <div class="col-md-4 pull-right hidden-xs hidden-sm"> <p class="text-right"><strong>ITSP.10.036</strong></p> </div> <!–MOBILE STARTS HERE–> <div class="hidden-lg hidden-md text-center"> <p><strong>September 2026 | Practitioner series</strong></p> </div> <!–pdf download–> <div class="col-md-12 mrgn-tp-lg"><div class="mrgn-bttm-md well well-sm col-md-4 col-sm-12 col-xs-12 pull-right mrgn-lft-md"> <p class="mrgn-tp-sm"><strong>Alternate format</strong>: <a href="/sites/default/files/itsp.10.036-e.pdf">Organizational cyber security and privacy risk management activities  – ITSP.10.036 (PDF, 1 MB)</a></p> </div></div> </div> <h2 class="text-info">Foreword</h2> <p>Organizational cyber security and privacy risk management activities (ITSP.10.036) is an unclassified publication issued under the authority of the Head, Canadian Centre for Cyber Security (Cyber Centre).</p> <p>This publication supersedes <abbr title="information technology">IT</abbr> security risk management: A lifecycle approach (ITSG-33), Annex 1 – Departmental <abbr title="information technology">IT</abbr> security risk management activities.</p> <p>For more information or to suggest amendments, contact the Cyber Centre:</p> <ul><li>by email: <a href="mailto:contact@cyber.gc.ca">contact@cyber.gc.ca</a></li> <li>by phone: <a href="tel:+16139497048">(613) 949-7048</a> or <span class="nowrap"><a href="tel:+18332923788">1-833-CYBER-88</a></span></li> </ul><h2 class="text-info">Effective date</h2> <p>This publication takes effect on September 14, 2026.</p> <h2 class="text-info">Revision history</h2> <ol><li><strong>First release:</strong> September 14, 2026</li> </ol><section><h2 class="text-info">Overview</h2> <p>This publication is part of a series of guidelines published by the Cyber Centre, under Cyber security and privacy risk management: A lifecycle approach. This publication describes a cyber security and privacy risk management process that includes activities at the organizational level.</p> <p>This publication is a tool to assist security and privacy practitioners in their efforts to protect organizations in compliance with applicable legislation, policies, directives and standards. It outlines a risk management framework designed to ensure that organizations effectively manage risks and obtain approval from the appropriate authorities on those risks.</p> <p>This publication focuses on 4 key objectives that guide organizations in effectively managing risks and ensuring compliance:</p> <ul><li>Identify and understand the cyber security and privacy needs of organizational programs and services and select a set of tailored controls and activities that meet these needs</li> <li>Allocate security controls that satisfy cyber security and privacy needs and the cyber security and risk management requirements of applicable legislation, policies, directives, and standards (for the Government of Canada (GC), namely Treasury Board of Canada Secretariat (TBS) policy instruments)</li> <li>Continuously monitor and assess the performance of organizational security and privacy controls and activities to detect incidents and identify vulnerabilities and deficiencies in a timely manner</li> <li>Update implemented controls and activities to respond to incidents, correct vulnerabilities, and continuously improve the cyber security and privacy posture of organizational systems</li> </ul><div class="panel panel-info"> <header class="panel-heading"><h3 class="panel-title"><abbr title="Government of Canada">GC</abbr>-specific note</h3> </header><div class="panel-body"> <p>In the <abbr title="Government of Canada">GC</abbr>, adherence to the Cyber security and privacy risk management: A lifecycle approach guidelines has many benefits for organizations, including:</p> <ul><li>compliance with the overall risk management strategy and objectives established by <abbr title="Treasury Board Secretariat">TBS</abbr></li> <li>addressing key aspects of cyber security and privacy</li> <li>managing cyber security and privacy risks, all in an efficient, consistent and cost-effective manner</li> </ul></div> </div> </section><section><details class="mrgn-tp-md"><summary><h2 class="h3">Table of contents</h2> </summary><ul class="list-unstyled lst-spcd"><li><a href="#1">1 Introduction</a> <ul class="lst-none"><li><a href="#1.1">1.1 Purpose</a></li> <li><a href="#1.2">1.2 Context in the Government of Canada</a></li> <li><a href="#1.3">1.3 Scope and applicability</a></li> <li><a href="#1.4">1.4 Audience</a></li> <li><a href="#1.5">1.5 Publication taxonomy</a></li> <li><a href="#1.6">1.6 Publication organization</a></li> </ul></li> <li><a href="#2">2 Cyber security and privacy risk management process</a></li> <li><a href="#3">3 Relationships with external processes</a></li> <li><a href="#4">4 Organizational cyber security and privacy risk management activities</a> <ul class="lst-none"><li><a href="#4.1">4.1 Define organizational cyber security and privacy needs</a></li> <li><a href="#4.2">4.2 Develop organizational security and privacy control and activity profiles</a></li> <li><a href="#4.3">4.3 Allocate security and privacy controls and activities</a></li> <li><a href="#4.4">4.4 Monitor and assess performance of security and privacy controls and activities</a></li> <li><a href="#4.5">4.5 Maintain authorization</a></li> <li><a href="#4.6">4.6 Update security and privacy controls and activities profile(s)</a></li> </ul></li> <li><a href="#5">5 Injury assessment and categorization process</a> <ul class="lst-none"><li><a href="#5.1">5.1 Concepts</a></li> <li><a href="#5.2">5.2 Injury assessment and categorization process</a></li> <li><a href="#5.3">5.3 Injury assessment and categorization process description</a></li> <li><a href="#5.4">5.4 Examples</a></li> </ul></li> </ul><h3>List of figures</h3> <ul class="list-unstyled"><li><a href="#fig1">Figure 1: Cyber security and privacy risk management process</a></li> <li><a href="#fig2">Figure 2: Organizational cyber security and privacy risk management activities</a></li> <li><a href="#fig3">Figure 3: Injury assessment and categorization process</a></li> </ul><h3>List of tables</h3> <ul class="list-unstyled"><li><a href="#tab1">Table 1: Examples of common controls and activities allocation</a></li> <li><a href="#tab2">Table 2: Examples of injury types and levels (<abbr title="Government of Canada">GC</abbr>-centric)</a></li> <li><a href="#tab3">Table 3: Examples of injury types and levels</a></li> <li><a href="#tab4">Table 4: Confidentiality – Non-national interest (Protected)</a></li> <li><a href="#tab5">Table 5: Confidentiality – National interest (Classified)</a></li> <li><a href="#tab6">Table 6: Integrity and availability</a></li> </ul><h3>List of annexes</h3> <ul class="list-unstyled"><li><a href="#ann.a">Annex A – Government of Canada injury levels conversion</a></li> </ul></details></section><section><h2 class="text-info" id="1">1 Introduction</h2> <p>The process described in this publication outlines a risk management framework designed to ensure that organizations effectively manage cyber security and privacy risks and obtain approval from the appropriate authorities for those risks. This publication also provides specifications on how to conduct injury assessment and categorization as part of risk management. The process described in this publication can be adapted to fit your organization’s needs.</p> <h3 id="1.1">1.1 Purpose</h3> <p>This publication is part of a series of guidelines published by the Cyber Centre, under Cyber security and privacy risk management: A lifecycle approach. It aims to support organizations in implementing a structured risk management approach. This publication outlines a set of activities to help organizations determine and achieve an appropriate level of security, aligned with business requirements and operational context. It also provides guidance on structuring a security assessment and authorization (SA&amp;A) framework, which formalizes how cyber security risks are assessed, documented, and managed across an organization. It describes how to conduct an injury assessment and categorization of the organization’s business processes and related information based on the potential impacts to confidentiality, integrity and availability (CIA).</p> <h3 id="1.2">1.2 Context in the Government of Canada</h3> <p>In the <a href="https://www.tbs-sct.canada.ca/pol/doc-eng.aspx?id=32611"><abbr title="Treasury Board Secretariat">TBS</abbr> Directive on Security Management (DSM)</a>, <abbr title="Treasury Board Secretariat">TBS</abbr> assigns to senior officials in the department’s security governance the responsibility for participating in and reporting to the department as well as assigning security responsibilities for programs, services and activities.</p> <p>The <abbr title="Treasury Board Secretariat">TBS</abbr> <abbr title="Directive on Security Management">DSM</abbr> instructs senior officials in a department’s security governance to identify security requirements and related resource needs for programs. It also advises them to ensure that security practices and security controls are defined, documented, implemented, monitored and maintained to meet identified security requirements for programs, services and activities. These senior officials shall document or recommend actions to be taken regarding residual risks for programs, services and activities. They need to establish processes to monitor, respond to, and report threats, vulnerabilities, security incidents and other security events. They are required to address security events that could impact programs, services and activities or that require an immediate or coordinated government-wide action. They must monitor and report on the effectiveness of security practices and controls and share the results with the chief security officer (CSO).</p> <p>The federal <a href="https://laws-lois.justice.gc.ca/eng/acts/P-21/"><em>Privacy Act</em></a>, along with the <a href="https://www.canada.ca/en/treasury-board-secretariat/services/access-information-privacy/privacy/privacy-policies-guidance.html"><abbr title="Treasury Board Secretariat">TBS</abbr> privacy governance suite</a>, assigns responsibility for ensuring compliance with legislative and policy requirements to senior officials and information custodians overseeing programs that collect and use personal information. Additionally, the responsibility for assessing compliance with these activities is assigned to the head of the government organization, although this role is often delegated in practice to a senior official.</p> <p>The management of cyber security risks overlaps with the management of privacy risks as data containing personal information is increasingly present in information systems throughout the data lifecycle. Privacy breaches can arise from cyber security incidents when there is a loss of confidentiality and/or integrity during data processing. This publication addresses the privacy risks as they relate to cyber systems only.</p> <p><a href="https://www.cyber.gc.ca/en/guidance/cyber-security-privacy-risk-management/itsp10033">Security and privacy controls and assurance activities catalogue (ITSP.10.033)</a> now includes privacy controls and activities to ensure that government institutions protect and manage personal information. These new additions help identify, assess, monitor and mitigate compliance gaps in programs and activities involving the collection, creation, retention, use, disclosure and disposal of personal information. Additionally, they aim to ensure that privacy is architected and then engineered into the design and operation of <abbr title="Government of Canada">GC</abbr> information systems.</p> <p>Organizations use the <a href="https://www.tbs-sct.canada.ca/pol/doc-eng.aspx?id=18309"><abbr title="Treasury Board Secretariat">TBS</abbr> Directive on Privacy Practices, Appendix C: Standard on Privacy Impact Assessment</a> to support this objective by ensuring that privacy implications are appropriately identified, assessed and resolved before a new or substantially modified program or activity involving personal information is implemented. The privacy impact assessment (PIA) is the component of risk management that focuses on ensuring compliance with the <em>Privacy Act</em> requirements. It also assesses the privacy implications of new or substantially modified programs and activities involving personal information. PIAs should be conducted in a manner that is commensurate with the privacy impact identified and respects the operating environment of the government institution.</p> <h3 id="1.3">1.3 Scope and applicability</h3> <p>This publication provides guidelines to organizations on the cyber security and privacy risk management activities that are performed by cyber security and privacy functions as part of an organizational cyber security program. System lifecycle cyber security and privacy risk management activities (ITSP.10.037) provides guidelines on the cyber security and privacy risk management activities that are performed at the system level.</p> <p>Adherence to these guidelines presents many benefits for organizations. These include adhering to established risk management strategy and objectives, addressing key aspects of cyber security and privacy in an efficient manner, and consistently and cost-effectively managing cyber security and privacy risks.</p> <div class="panel panel-info"> <header class="panel-heading"><h3 class="panel-title"><abbr title="Government of Canada">GC</abbr>-specific note</h3> </header><div class="panel-body"> <p>The guidelines in this publication apply to departments<sup id="fn1-rf"><a class="fn-lnk" href="#fn1"><span class="wb-inv">Footnote </span>1</a></sup> subject to the <a href="https://www.tbs-sct.canada.ca/pol/doc-eng.aspx?id=16578"><abbr title="Treasury Board Secretariat">TBS</abbr> Policy on Government Security (PGS)</a> that rely on information systems to support non-critical to critical departmental business programs, services, and activities in Unclassified, Protected, and Classified environments. Compliance with the requirements in the <abbr title="Policy on Government Security">PGS</abbr> and its supporting instruments is mandatory for designated federal departments.</p> </div> </div> <h3 id="1.4">1.4 Audience</h3> <p>This publication is intended to serve a diverse audience, including:</p> <ul><li>individuals with enterprise-level responsibilities, including enterprise architects and enterprise security architects</li> <li>individuals with system-development responsibilities, including mission or business owners, program managers, system engineers, system security engineers, privacy practitioners, hardware and software developers, system integrators, and acquisition or procurement officials or executives</li> <li>individuals with logistical or disposition-related responsibilities, including program managers, procurement officials or executives, system integrators, and property managers</li> <li>individuals with security and privacy implementation and operations responsibilities, including mission or business owners, system owners, information custodians, system administrators, continuity planners and system security or privacy officers</li> <li>individuals with security and privacy assessment and monitoring responsibilities, including auditors, system evaluators, control assessors, independent verifiers and validators and analysts</li> <li>security and privacy officials or executives (for example, chief information officers (CIOs), <abbr title="chief security officers">CSOs</abbr>, and chief privacy officers (CPOs))</li> <li>commercial entities, including industry partners, that produce component products and systems, create security and privacy technologies, or provide services or capabilities that support cyber security or privacy</li> </ul><div class="panel panel-info"> <header class="panel-heading"><h3 class="panel-title"><abbr title="Government of Canada">GC</abbr>-specific note</h3> </header><div class="panel-body"> <p>In the <abbr title="Government of Canada">GC</abbr>, this publication is intended for the audiences above, as well as for individuals who support departmental cyber security and privacy risk management activities, such as:</p> <ul><li>individuals with system, information security, privacy, or risk management and oversight responsibilities, including authorizing officials, <abbr title="chief information officers">CIOs</abbr>, <abbr title="chief security officers">CSOs</abbr>, senior officials in the department’s security governance, designated officials for cyber security, and appropriate privacy officials or executives</li> <li><abbr title="chief privacy officers">CPOs</abbr>, officials with delegated authority under the <em>Privacy Act,</em> and privacy practitioners</li> <li>individuals who participate in the definition, design, development, installation, and operation of information systems; more specifically, authorizers, project managers, cyber security architects, cyber security engineers, cyber security assessors, and members of cyber security operations groups</li> </ul></div> </div> <h3 id="1.5">1.5 Publication taxonomy</h3> <p>This publication is part of a series of guidelines that fall under "Cyber security and privacy risk management: A lifecycle approach." The documents in the series are as follows:</p> <ul><li>Overview – Cyber security and privacy risk management: A lifecycle approach (ITSP.10.035)</li> <li>Organizational cyber security and privacy risk management activities (ITSP.10.036)</li> <li>System lifecycle cyber security and privacy risk management activities (ITSP.10.037)</li> <li><a href="/en/guidance/cyber-security-privacy-risk-management/itsp10033/">Security and privacy controls and assurance activities catalogue (ITSP.10.033)</a></li> <li><a href="/en/guidance/cyber-security-privacy-risk-management/suggested-organizational-security-privacy-control-activity-profile-medium-impact-itsp10033-01">Suggested organizational security and privacy control and activity profile – Medium impact (ITSP.10.033-01)</a></li> <li>Assessment of security and privacy controls and assurance activities (ITSP.10.033-02)</li> </ul><h3 id="1.6">1.6 Publication organization</h3> <p>The remainder of this publication is organized as follows:</p> <ul><li><a href="#2">Section 2: Cyber security and privacy risk management process</a> introduces the cyber security and privacy risk management process at a high level and explains how it influences the system-level process presented in ITSP.10.037</li> <li><a href="#3">Section 3: Relationships with external processes</a> outlines how the cyber security and privacy risk management process interfaces with external processes</li> <li><a href="#4">Section 4: Organizational cyber security and privacy risk management activities</a> describes the detailed process of managing cyber security and privacy risks in an organization</li> <li><a href="#5">Section 5: Injury assessment and categorization process</a> presents the injury assessment and categorization process, including definitions of key concepts and illustrative examples</li> </ul></section><!–** TOP OF PAGE ******–><div class="clearfix"> </div> <div class="pull-right small text-muted mrgn-bttm-0"><a href="#wb-tphp">Top of page</a> <span aria-hidden="true" class="text-primary glyphicon glyphicon-circle-arrow-up"></span></div> <!–** END TOP OF PAGE **–> <section><h2 class="text-info" id="2">2 Cyber security and privacy risk management process</h2> <p>To manage cyber security and privacy risks efficiently and cost effectively, ITSP.10.036 and ITSP.10.037 describe an approach that organizations can adapt to fit their culture, mission and business objectives, business needs for security and privacy, and threats relevant to their activities.</p> <p><a href="#fig1">Figure 1</a> depicts the cyber security and privacy risk management process that is suggested in ITSP.10.036 and ITSP.10.037. The figure shows that cyber security and privacy risk management activities are occurring at 2 distinct levels in an organization: the organizational level and the system level.</p> <!– IMAGE EN –> <div class="panel panel-default"> <div class="panel-body"> <figure id="fig1"><figcaption class="mrgn-bttm-md text-center"><strong>Figure 1 – Cyber security and privacy risk management process</strong></figcaption><img alt="Flow chart demonstrating security and privacy risk management process – Long description immediately follows" class="img-responsive center-block mrgn-bttm-lg" src="/sites/default/files/images/itsp10036-fig1-e.png" /><details class="brdr-tp brdr-rght brdr-bttm brdr-lft mrgn-bttm-sm"><summary>Long description – Figure 1: Cyber security and privacy risk management process</summary><p>This figure describes the high-level organizational cyber security and privacy risk management activities. The figure also describes the system lifecycle cyber security and privacy risk management activities within projects and operational groups. It highlights how the cyber security and privacy risk management activities at both levels act together in a continuous cycle to efficiently maintain and improve the security posture of organizational systems.</p> <p>The <strong>Organizational cyber security and privacy risk management activities</strong> are intended for organizational cyber security and privacy authorities and encompass the following steps:</p> <ul class="list-unstyled"><li><strong>4.1 Define</strong> the organizational cyber security and privacy needs <ul><li>Submits organizational cyber security threat assessment report to Projects and operational groups in charge of system lifecycle cyber security and privacy risk management activities</li> </ul></li> <li><strong>4.2 Develop</strong> organizational security and privacy control and activity profiles <ul><li>Disseminated approved organizational security and privacy control and activity profiles</li> </ul></li> <li><strong>4.3 Allocate</strong> security and privacy controls and activities <ul><li>Lists of controls and activities allocations (common, hybrid, system-specific) are submitted to Projects and operational groups in charge of system lifecycle cyber security and privacy risk management activities</li> </ul></li> <li><strong>4.4 Monitor and assess</strong> performance of security and privacy controls and activities</li> <li><strong>4.5 Maintain</strong> authorization</li> <li><strong>4.6 Update</strong> security and privacy control and activity profile(s) <ul><li>Then return to step 2, Develop organizational security and privacy control and activity profiles</li> </ul></li> </ul><p>The <strong>System lifecycle cyber security and privacy risk management activities</strong> enacted by projects and operational groups are aimed at project managers, security and privacy architects, assessors and engineers, and developers, among others. They encompass the following steps:</p> <ol><li>Concept: Define system cyber security and privacy needs</li> <li>Development: Architect and design secure systems</li> <li>Production: Implement, integrate, verify and validate secure systems</li> <li>Operations and support: Operate, monitor and maintain secure systems <ul><li>Submit Cyber security and privacy performance feedback to Organizational cyber security and privacy authorities to inform step 4, Monitor and assess performance of security and privacy controls and activities</li> </ul></li> <li>Retirement: Securely dispose of assets at retirement</li> </ol></details></figure></div> </div> <!– END IMAGE EN –> <p>At the organizational level, which is the focus of this publication, cyber security and privacy risk management activities are conducted by the organization’s cyber security and privacy functions (refer to <a href="#4">section 4</a>). To successfully carry out this process, organizational cyber security and privacy programs need to collaborate closely as their objectives can coincide and be complementary. The objectives of these activities are to:</p> <ul><li>define organizational cyber security and privacy needs and associated controls and activities</li> <li>allocate responsibility for security and privacy controls and activities</li> <li>continuously monitor and assess the performance of allocated controls and activities</li> <li>identify required updates to controls and activities (changes to existing controls/activities or addition of controls/activities) based on the results of the continuous monitoring, assessment, and authorization maintenance activities, and oversee the implementation of these updates</li> </ul><p>At the system level, cyber security and privacy risk management activities are conducted by information technology and operational technology (IT/OT) projects and <abbr title="information technology and operational technology">IT/OT</abbr> operations groups. They follow the implementation, operation and disposal phases of systems, as outlined in System lifecycle cyber security and privacy risk management activities (ITSP.10.037). The objectives of the system-level activities are to:</p> <ul><li>define cyber security and privacy needs as well as associated controls and activities for systems based on the activities of the cyber security and privacy functions</li> <li>design, develop, or acquire systems that satisfy defined controls and activities</li> <li>integrate, test, and install systems with security and privacy</li> <li>operate, monitor, and maintain the security and privacy of systems during the operations and maintenance phase</li> <li>securely dispose of assets when systems are retired</li> <li>securely dispose of personal information assets when they no longer are required by the organization</li> </ul></section><section><h2 class="text-info" id="3">3 Relationships with external processes</h2> <p>To fully benefit from this publication, organizations need to be aware of the following important relationships between the organizational cyber security and privacy risk management activities and other processes:</p> <ul><li><strong>Integrated risk management</strong> <ul><li>Organizational cyber security and privacy risk management activities can support organizational integrated risk management, as defined by the <a href="https://www.tbs-sct.canada.ca/pol/doc-eng.aspx?id=19422"><abbr title="Treasury Board Secretariat">TBS</abbr> Framework for the Management of Risk</a></li> <li>These activities form a continuous, proactive, and systematic process to understand, manage, and communicate cyber security and privacy risks from an organization-wide perspective</li> </ul></li> <li><strong>Business impact analysis</strong> <ul><li>Business impact analyses provide valuable insights for identifying an organization’s security and privacy requirements and for assessing the sensitivity and criticality of its business activities</li> <li>Together, these factors determine the appropriate security category for those activities</li> </ul></li> <li><strong>Enterprise architecture</strong> – An enterprise architecture function provides key inputs to help ensure that organization-wide cyber security and privacy requirements are taken into consideration when defining, allocating and updating controls and activities.</li> </ul><div class="panel panel-info"> <header class="panel-heading"><h3 class="panel-title"><abbr title="Government of Canada">GC</abbr>-specific note</h3> </header><div class="panel-body"> <p>In the <abbr title="Government of Canada">GC</abbr>, another process needs to be considered: the <strong>algorithmic impact assessment (AIA)</strong>. This is a mandatory risk assessment tool intended to support the <a href="https://www.tbs-sct.canada.ca/pol/doc-eng.aspx?id=32592"><abbr title="Treasury Board Secretariat">TBS</abbr> Directive on Automated Decision-Making</a>. It produces a risk value that reflects the potential risks of the automated decision system, and departments and agencies should consider this score as part of their risk management activities.</p> </div> </div> </section><!–** TOP OF PAGE ******–><div class="clearfix"> </div> <div class="pull-right small text-muted mrgn-bttm-0"><a href="#wb-tphp">Top of page</a> <span aria-hidden="true" class="text-primary glyphicon glyphicon-circle-arrow-up"></span></div> <!–** END TOP OF PAGE **–> <section><h2 class="text-info" id="4">4 Organizational cyber security and privacy risk management activities</h2> <p>This section describes organizational cyber security and privacy risk management activities that the Cyber Centre recommends be incorporated into an organizational security and privacy program. It also provides guidelines on how to effectively complete them.</p> <div class="panel panel-info"> <header class="panel-heading"><h3 class="panel-title"><abbr title="Government of Canada">GC</abbr>-specific note</h3> </header><div class="panel-body"> <p>Before incorporating ITSP.10.036 activities in their departmental security and privacy program, departments should ensure that their governance structure (roles, responsibilities, and decision-making authorities) aligns with the governance structure found in the latest <abbr title="Treasury Board Secretariat">TBS</abbr> policy instruments.</p> </div> </div> <p><a href="#fig2">Figure 2</a> <strong>Organizational cyber security and privacy risk management activities</strong> depicts the activities of the cyber security and privacy risk management process that are conducted at the organizational level. The main goal of these activities is to allocate and maintain a set of security and privacy controls and activities that are tailored to the specific security and privacy needs and objectives of an organization.</p> <p>The protection of personal information<sup id="fn2-rf"><a class="fn-lnk" href="#fn2"><span class="wb-inv">Footnote </span>2</a></sup> is a shared and multifaceted responsibility requiring concerted and coordinated efforts between the security and privacy functions throughout the risk management process.</p> <p>Key outputs of organizational cyber security and privacy risk management activities are control and activities profiles, and organizational threat repositories both of which serve as key inputs to the information system level of the risk management process for the deployment of controls and activities in information systems.</p> <!– IMAGE EN –> <div class="panel panel-default"> <div class="panel-body"> <figure id="fig2"><figcaption class="mrgn-bttm-md text-center"><strong>Figure 2 – Organizational cyber security and privacy risk management activities</strong></figcaption><img alt="Flow chart demonstrating organizational cyber security and privacy risk management activities – Long description immediately follows" class="img-responsive center-block mrgn-bttm-lg" src="/sites/default/files/images/itsp10036-fig2-e.png" /><details class="brdr-tp brdr-rght brdr-bttm brdr-lft mrgn-bttm-sm"><summary>Long description – Figure 2: Organizational cyber security and privacy risk management activities</summary><p>The figure presents a horizontal process flow diagram titled "Organizational cyber security and privacy risk management activities." There is a dark header that spans the top and states the target audience as organizational cyber security and privacy authorities. Along the left edge of the figure, a vertical side-bar says "Organizational cyber security and privacy functions."</p> <p>Across the center, there are six yellow rectangles representing the main organizational activities, arranged left to right and connected by right-pointing arrows. The activities are:</p> <ul class="list-unstyled"><li><strong>4.1 Define</strong> organizational cyber security and privacy needs</li> <li><strong>4.2 Develop</strong> organizational security and privacy control and activity profiles</li> <li><strong>4.3 Allocate</strong> security and privacy controls and activities</li> <li><strong>4.4 Monitor and Assess</strong> performance of security and privacy controls and activities</li> <li><strong>4.5 Maintain</strong> authorization</li> <li><strong>4.6 Update</strong> security and privacy controls and activities profile(s)</li> </ul><p>Each of the first 3 activities has a downward arrow to a white "key deliverable" box beneath it:</p> <ul><li>Under <strong>4.1</strong>: Organizational cyber security threat assessment report</li> <li>Under <strong>4.2</strong>: Disseminated approved organizational security and privacy control and activity profiles</li> <li>Under <strong>4.3</strong>: Lists of controls and activities allocation (common, hybrid, system-specific)</li> </ul><p>Under both <strong>4.1</strong> and <strong>4.3</strong>, there is a downward pointing arrow with a caption line that reads "Key inputs to the security and privacy engineering process."</p> <p>A large white upward-pointing arrow is positioned beneath the <strong>4.4</strong> section. The arrow is labeled "Cyber security and privacy performance feedback" and signifies that monitoring results feed back into the organizational process. To the right of the upward arrow, a note says "Output from system security and privacy control monitoring," to indicate that system-level monitoring produces feedback used by the organization to reassess and update, when necessary, its profiles and control allocations.</p> <p>The final activity, <strong>4.6 Update</strong>, is connected back toward the earlier step <strong>4.2 "Develop organizational security and privacy control and activity profiles"</strong> via a looping arrow along the top of the flow.</p> </details></figure></div> </div> <!– END IMAGE EN –> <p>The organizational cyber security and privacy risk management activities seek to achieve the following objectives:</p> <ul><li>Identify and document the business needs for security and privacy of organizational business activities</li> <li>Conduct and maintain organizational threat assessments that security and privacy authorities and practitioners can leverage to identify relevant threat actors, inform risk and robustness determinations throughout the development lifecycle, and produce more consistent outcomes</li> <li>Develop organizational controls and activities profiles tailored to business needs for security and privacy <ul><li>Leverage profile(s) in projects that implement or update organizational information systems</li> </ul></li> <li>Allocate the security and privacy controls and activities within the enterprise security architecture</li> <li>Define which controls and activities are common, hybrid or system-specific to support the assignment of responsibilities across the enterprise</li> <li>Coordinate between organizational security and privacy functions (for example, physical security and personnel security) to ensure a consistent approach to building and operating dependable information systems</li> <li>Monitor and assess the performance of implemented controls and activities in protecting organizational business activities Devise and implement corrective measures to improve performance and the organization’s overall security and privacy posture</li> </ul><h3 id="4.1">4.1 Define organizational cyber security and privacy needs</h3> <p>This section outlines the process for defining organizational cyber security and privacy needs. It includes the following sub-activities:</p> <ul><li>Define the scope of the organization’s cyber security and privacy risk assessment activities (<a href="#4.1.1">section 4.1.1</a>)</li> <li>Identify the business needs for security and privacy of organizational business activities (<a href="#4.1.2">section 4.1.2</a>)</li> <li>Perform injury assessment and categorization of organizational business activities (<a href="#4.1.3">section 4.1.3</a>)</li> <li>Select the organizational cyber security threat and risk assessment (TRA) methodology (<a href="#4.1.4">section 4.1.4</a>)</li> <li>Select the organizational privacy assessment methodology (<a href="#4.1.5">section 4.1.5</a>)</li> <li>Conduct an organizational cyber security threat assessment (<a href="#4.1.6">section 4.1.6</a>)</li> </ul><p>Each of these activities aiming to define organizational cyber security and privacy needs is described in the sections that follow. These activities are performed when first implementing the guidelines presented in this publication, and then as required based on the assessment of the security and privacy posture of an organization’s information systems (refer to <a href="#4.6">section 4.6</a>).</p> <h4 id="4.1.1">4.1.1 Define the scope of the cyber security and privacy risk assessment activities</h4> <p><strong>Objective:</strong> Define the scope of the organization’s cyber security and privacy risk assessment activities.</p> <p><strong>Suggested related roles:</strong> Senior security officials (for example, <abbr title="chief security officer">CSO</abbr>, chief information security officer (CISO)), senior privacy officials, cyber security advisors, enterprise security architects, privacy advisors, privacy architects, project directors</p> <p>The scope can be characterized by the:</p> <ul><li>organization’s programs, services, information, and business activities requiring protection</li> <li>organization’s programs, services, and activities involving the creation, collection, and handling of personal information that are under its control</li> <li>major organizational assets (for example, business applications, information systems, data centres, local areas networks, data processed and stored, information) and their geographical locations</li> <li>core technologies that are used in organizational information systems</li> <li>procurement strategies (for example, internal, outsourced)</li> </ul><p>The scope should clearly delineate the organizational business activities and assets that are within its boundaries, and those that are excluded and why. The scope should also identify external dependencies like external service providers.</p> <p><strong>Output:</strong> Definition of the scope of the organization’s cyber security and privacy risk assessment activities and personal information inventory.</p> <h4 id="4.1.2">4.1.2 Identify business needs for security and privacy</h4> <p><strong>Objective:</strong> Identify the business needs for security and privacy of organizational business activities.</p> <p><strong>Suggested related roles:</strong> Senior security officials, such as <abbr title="chief security officers">CSOs</abbr> and <abbr title="chief information security officers">CISOs</abbr>, senior privacy officials, cyber security advisors, privacy advisors, project directors</p> <p>Within the context of cyber security and privacy risk management, business needs for security and privacy refer to the need to protect organizational business activities and information against adverse events that could affect the ability of organizations to meet their mission, objectives, and obligations. Legislation or regulations that impose protection requirements for information systems and personal information, create business needs for security and privacy to ensure compliance. Business needs for security and privacy, in turn, shape formal security and privacy objectives. They may originate from:</p> <ul><li>requirements found in legislation, regulations, policies, directives, standards, jurisprudence and objectives governing organizational business activities and information management (for example, in the <abbr title="Government of Canada">GC</abbr>, a departmental results framework)</li> <li>generally or organizationally recognized information system threat exposures and compliance obligations</li> <li>contractual or negotiated obligations (for example, contracts, memoranda of understanding (MOUs), service level agreements (SLAs) and information sharing agreements (ISAs))</li> </ul><p>Business needs for security include privacy-related requirements needing the support of cyber security for privacy risk management purposes.</p> <p>Examples of business needs for security and privacy include the need to:</p> <ul><li>limit access to sensitive program information or personal information to authorized individuals for authorized actions</li> <li>limit secondary uses or disclosures of personal information</li> <li>verify the identity of citizens before disclosing personal information that relates to them</li> <li>ensure that only authorized individuals can approve financial payment to a recipient</li> </ul><p>Organizations should identify business needs for security and privacy with the help of business analysts, system owners, system architects and security/privacy architects. Useful inputs to this activity include business process documentation (for example, business process descriptions, use cases, concepts of operations (CONOPS)) and impact assessments (for example, business impact analysis, <abbr title="privacy impact assessment">PIA</abbr>, algorithmic impact assessment).</p> <p><strong>Output:</strong> Document stating the business needs for security and privacy for the organization’s business activities.</p> <h4 id="4.1.3">4.1.3 Perform injury assessment and categorization of organizational business activities</h4> <p><strong>Objective:</strong> Perform the assessment and categorization of injuries to organizational business activities.</p> <p><strong>Suggested related roles:</strong> Security advisors, privacy advisors, project director</p> <p>An injury category expresses the highest levels of expected injuries from threat compromise with respect to privacy and to the security objectives of confidentiality, integrity, and availability. Safety requirements may elevate the categorization for integrity and availability.</p> <p>Business activities are categorized by first determining the expected injuries from a cyber-related threat compromise to the national and non-national interests that the business activities serve and then determining the level of these expected injuries. While the breach of personal information may result in varying levels of injury, the compliance obligations do not change based on the type of the personal information. The comprehensive security categorization of the information should include consideration of the presence of personal information.</p> <p>Organizations can categorize their business activities following the process specified in <a href="#5">section 5</a>.</p> <p><strong>Output:</strong> Injury categorization report for organizational business activities.</p> <h4 id="4.1.4">4.1.4 Select organizational cyber security threat and risk assessment methodology</h4> <p><strong>Objective:</strong> Select the methodology by which security and privacy threats and risks will be assessed across the organization.</p> <p><strong>Suggested related roles:</strong> Senior security officials (<abbr title="chief security officer">CSO</abbr>, <abbr title="chief information security officer">CISO</abbr>), security advisors</p> <p>A cyber security <abbr title="threat risk assessment">TRA</abbr> methodology must be selected at this stage, as an organization will use the threat assessment portion when assessing threats to organizational business activities. In addition, projects will use this methodology when performing TRA activities.</p> <p>An example of a commonly used <abbr title="threat risk assessment">TRA</abbr> methodology in the <abbr title="Government of Canada">GC</abbr> is the <a href="https://www.cyber.gc.ca/en/tools-services/harmonized-tra-methodology">CSE-<abbr title="Royal Canadian Mounted Police">RCMP</abbr> Harmonized Threat and Risk Assessment (HTRA) Methodology</a>.</p> <p><strong>Output:</strong> Selection of organizational cyber security <abbr title="threat risk assessment">TRA</abbr> methodology.</p> <h4 id="4.1.5">4.1.5 Select the organizational privacy assessment methodology</h4> <p><strong>Objective:</strong> Select the methodology by which privacy risks and compliance gaps will be assessed across the organization.</p> <p><strong>Suggested related roles:</strong> Senior privacy officials, privacy advisors</p> <p>At this stage, a privacy assessment methodology needs to be selected to identify, assess, and mitigate potential privacy risks associated with new or modified programs, projects, or technologies.</p> <p>In the <abbr title="Government of Canada">GC</abbr>, privacy assessments follow <abbr title="Treasury Board Secretariat">TBS</abbr> policy, beginning with the completion of a privacy checklist template. This checklist helps determine whether an initiative requires a privacy protocol, a <abbr title="privacy impact assessment">PIA</abbr>, or no further privacy assessment. Each assessment methodology is built into its corresponding template, so selecting the appropriate template also determines the assessment approach.</p> <p><strong>Output:</strong> Selection of organizational privacy assessment methodology. In the <abbr title="Government of Canada">GC</abbr>, an additional output will be the completion of a privacy checklist that determines what template any additional privacy analysis should use.</p> <h4 id="4.1.6">4.1.6 Conduct an organizational cyber security threat assessment</h4> <p><strong>Objective:</strong> Conduct an initial organization-wide cyber security threat assessment that will guide the selection of security and privacy controls and activities, and that projects will leverage when implementing systems.<sup id="fn3-rf"><a class="fn-lnk" href="#fn3"><span class="wb-inv">Footnote </span>3</a></sup></p> <p><strong>Suggested related roles:</strong> Intelligence analysts, cyber security threat analysts</p> <p>This activity will identify and qualify threats of relevance to the in-scope organizational business activities.</p> <p>The cyber security threat assessment serves 2 purposes:</p> <ul><li>informing the <abbr title="threat risk assessment">TRA</abbr></li> <li>supporting threat context decisions</li> </ul><p>Organizations may specify a subset of all potential threats from which they want to protect their business activities. This implies that some threats may be identified and considered but be deemed out of scope for various reasons. For example, an organization may find that protecting against a threat would be too costly or complex, or that the protection would excessively limit a business activity’s supporting functionality. Threat information, including decisions and justification for excluding specific threats, is documented in an organizational threat context.</p> <p>An organization-wide threat assessment is a valuable tool that can be used to select, tailor, allocate, update, and improve the implemented security and privacy controls and activities. Its results, along with organizational business needs for security and privacy, form a strong foundation for developing organizational control and activity profiles.</p> <p>When developing organizational control and activity profiles, more focused, domain-specific threat repositories may be created to document detailed information on threats relevant to specific business domains.</p> <p>An effective organizational threat assessment should evaluate and consider:</p> <ul><li>key organizational business activities</li> <li>the injury assessment associated with these activities</li> <li>relevant threats to the business</li> <li>general exposures that could impact business activities (for example, physical locations vulnerable to earthquakes)</li> </ul><p>In the <abbr title="Government of Canada">GC</abbr>, departmental threat assessments are best conducted by multidisciplinary teams with the assistance of the CSO’s office, lead <abbr title="Government of Canada">GC</abbr> security agencies, and privacy practitioners.</p> <p><strong>Output:</strong> Organizational threat assessment repository documenting the threats and exposures of relevance to the business.</p> <!–** TOP OF PAGE ******–> <div class="clearfix"> </div> <div class="pull-right small text-muted mrgn-bttm-0"><a href="#wb-tphp">Top of page</a> <span aria-hidden="true" class="text-primary glyphicon glyphicon-circle-arrow-up"></span></div> <!–** END TOP OF PAGE **–> <h3 id="4.2">4.2 Develop organizational security and privacy control and activity profiles</h3> <p><strong>Objective:</strong> Create organizational security and privacy control and activity profiles that are tailored to security and privacy needs.</p> <p><strong>Suggested related roles:</strong> Senior security officials, such as <abbr title="chief security officers">CSOs</abbr> and <abbr title="chief information security officers">CISOs</abbr>, senior privacy officials, enterprise security architects, privacy architects, security advisors, privacy advisors</p> <p>Depending on their business scope (for example, managing a single initiative, overseeing multiple projects, or operating in multiple security domains) and the complexity of their business activities, organizations may maintain a single enterprise-wide profile or several domain-specific profiles tailored to different areas of their business.</p> <p>The cyber security and privacy functions of an organization are expected to collaborate on developing profiles since privacy risks may arise from unauthorized system activities. However, not all privacy risks associated with the handling of personal information in an information system falls within the scope of information security. Privacy risks can result from activities related to the creation, collection, retention, accuracy, use, disclosure, or disposition of personal information under the control of an organization. Hence, the privacy function of an organization will likewise select, allocate, monitor, and assess privacy controls in order to comply with legislative requirements.</p> <p>Organizational profiles are best developed with the support of several key organization-wide information management processes. These enterprise processes complement the cyber security and privacy risk management process and can assist organizations in appropriately select and tailor security and privacy controls and activities. The supporting processes are:</p> <ul><li>the review of the organizational business activities and their security and privacy needs</li> <li>the prioritization of organizational business activities according to strategic goals related to availability and business continuity (in general terms, this means ensuring that the most critical activities will continue or remain available)</li> <li>the definition of the types of information assets needed to successfully execute the departmental business activities, their criticality and sensitivity, and their internal and external flows</li> <li>the creation of data dictionaries that document the collections of personal information within the organization</li> <li>the definition of a conceptual enterprise architecture that includes cyber security and privacy requirements</li> </ul><p>There are 4 steps to developing organizational security and privacy control and activity profiles:</p> <ol><li>Define business domains</li> <li>Define security architecture approaches</li> <li>Select security and privacy controls and activities</li> <li>Approve the organizational security and privacy control and activity profiles</li> </ol><p>These 4 steps are described in detail in the subsections that follow.</p> <h4 id="4.2.1">4.2.1 Define business domains</h4> <p><strong>Objective:</strong> Define the business domains of an organization in support of developing the required organizational security and privacy control and activity profiles.</p> <p><strong>Suggested related roles:</strong> Enterprise security architects, privacy architects</p> <p>A business domain is characterized by the security categories of its business activities and their relevant cyber security threats. Therefore, business domains may have differing protection needs, leading to different profiles.</p> <p>For example, consider a set of business activities involving the distribution of non-sensitive publications and a second set of business activities involving high-value, critical financial transactions. In the latter scenario, the financial activities would likely have a higher injury category and face more significant threats. This analysis would lead to the definition of 2 domains requiring 2 different domain profiles.</p> <p>Organizations have some flexibility in how they define their business domains. They should document in detail how these domains are defined, the injury categories of the business activities, and the significance of the threat environment. Note that the organizational business activities should have been defined earlier (in whole or in part) during the injury assessment and categorization activity (see <a href="#4.1.3">section 4.1.3</a>).</p> <p>This activity produces business domain definitions. Each defined business domain should include</p> <ul><li>a description of the business domain’s business objectives, processes and information assets</li> <li>the injury category of the business domain</li> <li>a characterization of the threat environment of relevance to the business domain (that is, the highest deliberate threat (Td) level to be mitigated)</li> <li>a statement of the highest level of risk that the business owner deems acceptable</li> </ul><p><strong>Output:</strong> Business domain definitions, including the objectives, processes and information assets, the injury category, the threat environment, and a statement of the level of risk deemed acceptable.</p> <h4 id="4.2.2">4.2.2 Define security architecture approaches</h4> <p><strong>Objective:</strong> Define a set of security architecture approaches that align with the organizational cyber security and privacy philosophy, culture, objectives, and priorities.</p> <p><strong>Suggested related roles:</strong> Enterprise security architects, privacy architects</p> <p>At the organizational level, security architecture approaches influence the selection of certain technical security controls and activities related to the security engineering principles applied in the approach. At the information system level, security architecture approaches play an equally important role as a guide for tailoring security controls in profiles to the specific security needs of information systems and for specifying appropriate security designs.</p> <p>Enterprise security architects and privacy architects will likely seek approval from senior officials for their chosen security architecture approaches, as these decisions often have budget and capability implications.</p> <p>When defining security architecture approaches, organizations should take into consideration the following inputs:</p> <ul><li>business needs for security and privacy (<a href="#4.1.2">section 4.1.2</a>)</li> <li>injury assessment and categorization of organizational business activities (<a href="#4.1.3">section 4.1.3</a>)</li> <li>threats documented in the organizational threat assessment (<a href="#4.1.6">section 4.1.6</a>)</li> <li>generally accepted cyber security principles and best practices such as <ul><li>Sherwood Applied Business Security Architecture’s (SABSA) <a href="https://sabsa.org/other-resources/">Enterprise Security Architecture – A Business-Driven Approach</a></li> <li>Cyber Centre’s <a href="https://www.cyber.gc.ca/sites/default/files/cyber/publications/itsp80022-e.pdf">Baseline security requirements for network security zones (ITSP.80.022 v2.0)</a></li> <li>National Institute of Standards and Technology’s (NIST) <a href="https://csrc.nist.gov/pubs/sp/800/160/v1/r1/final">Special Publication (SP) 800-160 Vol.1 Rev.1 Engineering Trustworthy Secure Systems</a></li> <li><abbr title="National Institute of Standards and Technology">NIST</abbr>’s <a href="https://csrc.nist.gov/pubs/sp/800/160/v2/r1/final">SP 800-160 Vol. 2 Rev. 1 Developing Cyber-Resilient Systems: A Systems Security Engineering Approach</a></li> </ul></li> <li>generally accepted privacy principles and best practices such as <ul><li>Fair Information Principles</li> <li>Generally Accepted Privacy Principles (GAPPs)</li> </ul></li> </ul><p>The following are examples of security architecture approaches that influence the selection of security and privacy controls and activities and the specification of security designs that organizations can use:</p> <ul><li>Holistic versus piecemeal approach to design, where a holistic approach considers security and privacy at the various layers of an information system (for example, application, data management, middleware, platform, and communication layers) and includes an understanding of information flows between the various layers</li> <li>Defence-in-depth/layered defence versus focus on 1-layer protection</li> <li>Eggshell (hard perimeter/soft centre) versus honeycomb versus end-point protection approaches to design</li> <li>Appropriate security modes of operation (dedicated, system-high, compartmented, multilevel, multiple independent levels of security (MILS))</li> <li>Tiered application architecture (presentation, business logic, data repositories) applying security zoning versus monolithic system architecture</li> <li>Use of common security services with standardized application programming interface (for example, organizational authentication system) versus implementing custom security services</li> <li>Infrastructure security versus data asset security (for example, digital rights management)</li> <li>Societal considerations (for example, employees’ privacy versus monitoring) versus enforcement of hard-to-use but more secure controls</li> </ul><p><strong>Output:</strong> List of selected security architecture approaches and justifications.</p> <h4 id="4.2.3">4.2.3 Select organizational security and privacy controls and activities</h4> <p><strong>Objective:</strong> Select organizational security and privacy controls and activities.</p> <p><strong>Suggested related roles:</strong> Enterprise security architects, privacy architects, security advisors, privacy advisors</p> <p>Organizations develop a control and activity profile for each of their defined business domains. If an organization defines several business domains, then it can develop the profiles gradually, starting with one that is urgently needed, and then developing the others over time.</p> <p>There are 3 approaches to implementing controls and activities:</p> <ul><li>a common (inheritable) implementation approach</li> <li>a system-specific implementation approach</li> <li>a hybrid implementation approach</li> </ul><p>The implementation approaches define the scope of applicability for the control or activity, its shared nature or inheritability, and the responsibility for control or activity development, implementation, assessment, and authorization. Each implementation approach has a specific objective and focus that helps organizations select the appropriate controls and activities, effectively implement them, and satisfy security and privacy requirements. A specific implementation approach may save money by leveraging security and privacy capabilities across multiple systems and operational environments.</p> <p>A common control or activity is deployed as a shared service to the organization. Implementing common controls or activities results in a capability that is inheritable by multiple systems or programs. System-specific controls and activities are the primary responsibility of the system owner. Organizations can implement a hybrid control or activity if one part of the control or activity is common (inheritable) and the other part is system-specific. For example, an organization may implement contingency planning control CP-02 using a predefined template for all organizational information systems, with individual system owners tailoring the plan for system-specific uses where appropriate. For privacy, an organization-wide policy or procedure would be an example of a common privacy control or activity. For more information, refer to <a href="/en/guidance/cyber-security-privacy-risk-management/itsp10033/concepts-structure#2.3">Security and privacy controls and assurance activities catalogue (ITSP.10.033) section 2 Implementation approaches</a>.</p> <p>Organizational control and activity profiles are used by the cyber security and privacy functions to coordinate the allocation of common controls and activities to the enterprise architecture. They also inform projects about the controls and activities their system inherits, as well as those they must implement to secure the system they are developing or updating.</p> <p>When developing organizational control and activity profiles, organizations select controls and activities from the ITSP.10.033 catalogue, and tailor them to satisfy organizational security and privacy needs (<a href="#4.1.2">section 4.1.2</a>). The selection and tailoring of controls and activities is guided by the results of their organizational privacy assessment (<a href="#4.1.5">section 4.1.5</a>), organizational cyber security threat assessment (<a href="#4.1.6">section 4.1.6</a>) and the defined security architecture approaches (<a href="#4.2.2">section 4.2.2</a>). For organizations that have an enterprise architecture function, security and privacy practitioners should also consider cyber security-related artefacts when selecting and tailoring system-specific controls and activities.</p> <p>Organizational profiles should also document the business context and assumptions under which they were developed by describing:</p> <ul><li>in-scope business activities and related business needs for security and privacy</li> <li>injury assessment and categorization of in-scope business activities</li> <li>the threat context</li> <li>legislative and policy compliance artefacts</li> <li>defined security architecture approaches</li> <li>any other technical constraints or assumptions that might influence the selection of controls and activities</li> </ul><p>A key input to the control and activity profile development process is the organizational threat repository (see <a href="#4.1.6">section 4.1.6</a>). When developing domain profiles, organizations may refine organizational threat definitions based on additional threat information that is specific to each domain’s business activities. When this occurs, organizations can document these refined threat definitions in domain-specific threat repositories.</p> <p>Several of the controls and activities in the ITSP.10.033 catalogue should be considered for allocation as common controls and activities.</p> <p>Examples of common security and privacy controls and activities are:</p> <ul><li>an organizational personnel security screening program supporting the screening of cyber security personnel</li> <li>a physical security program supporting the protection of technological facilities</li> <li>security incident and privacy breach management performed as part of cyber security and privacy functions to provide a comprehensive situational awareness</li> <li>a system operated by an <abbr title="information technology and operational technology">IT/OT</abbr> operations group that provides a common end-user authentication solution</li> <li>an organization-wide electronic log monitoring system, operated by a team from the cyber security function, which provides separation of duties between the <abbr title="information technology and operational technology">IT/OT</abbr> operations and the cyber security function</li> <li>a cyber security awareness and training program administered by the organizational learning centre</li> </ul><p>To develop their organizational profile, organizations may leverage the medium impact profile in ITSP.10.033-01, which is based on the ITSP.10.033 catalogue.</p> <p><strong>Output:</strong> An organization-wide security and privacy control and activity profile or a set of domain security and privacy control and activity profiles.</p> <h4 id="4.2.4">4.2.4 Approve the organizational profile(s)</h4> <p><strong>Objective:</strong> Approve the organizational security and privacy control and activity profile(s).</p> <p><strong>Suggested related roles:</strong> Senior security officials, senior privacy officials</p> <p>Once the security and privacy control and activity profile(s) have been completed, senior officials in the organization’s security and privacy governance should review them and seek approval from organizational authorities (for example, <abbr title="chief executive officer">CEO</abbr>, <abbr title="chief information officer">CIO</abbr>, <abbr title="chief security officer">CSO</abbr>, <abbr title="chief information security officer">CISO</abbr>, <abbr title="chief privacy officer">CPO</abbr>, deputy head) as required. As part of this process, these senior officials should also ensure that the organizational controls and activities specified in the profiles satisfy organizational security and privacy needs, and that they adequately address threats and privacy compliance obligations. They should ensure that there will be a good balance between the implementation of controls and activities and the levels of residual risk that the organization is ready to assume.</p> <p><strong>Output:</strong> Approved organizational security and privacy control and activity profile(s).</p> <!–** TOP OF PAGE ******–> <div class="clearfix"> </div> <div class="pull-right small text-muted mrgn-bttm-0"><a href="#wb-tphp">Top of page</a> <span aria-hidden="true" class="text-primary glyphicon glyphicon-circle-arrow-up"></span></div> <!–** END TOP OF PAGE **–> <h3 id="4.3">4.3 Allocate security and privacy controls and activities</h3> <p><strong>Objective:</strong> Coordinate the allocation of security and privacy controls and activities.</p> <p><strong>Suggested related roles:</strong> Enterprise security architects, privacy architects</p> <p>Following the approval of organizational security and privacy control and activity profiles, senior officials in the organization’s security and privacy governance and their staff can proceed with coordinating the allocation of controls and activities.</p> <p>Developing a strategy for allocating security and privacy controls and activities starts with determining which ones are common, hybrid, or system-specific. It is also important to identify the responsible parties for delivering and operating common and hybrid controls and activities within the organization. Engaging an enterprise architecture board can help facilitate this process.</p> <h4 id="4.3.1">4.3.1 Allocate common controls and activities</h4> <p><strong>Objective:</strong> Coordinate and document the allocation of common controls and activities.</p> <p><strong>Suggested related roles:</strong> Enterprise security architects, privacy architects</p> <p>In an established organization, certain controls and activities may already be delivered through shared or common services. These should be identified and documented within the enterprise security architecture. For controls and activities that are not yet in place, responsibilities should be clearly defined and allocated as part of the enterprise security architecture function.</p> <p>Organizations have several options to allocate common security and privacy controls and activities. They can be allocated:</p> <ul><li>as part of the cyber security and privacy functions</li> <li>to an organizational <abbr title="information technology and operational technology">IT/OT</abbr> operations group</li> <li>to other groups within the organization, or outsourced to another organization or service provider</li> <li>to a new shared or common service</li> </ul><p><a href="#tab1">Table 1</a> provides examples of how common security and privacy controls and activities can be allocated.</p> <div class="table-responsive"> <table class="table table-bordered"><caption class="text-center" id="tab1">Table 1: Examples of common controls and activities allocation</caption> <thead><tr class="active"><th class="text-center" scope="col">Common controls and activities</th> <th class="text-center" scope="col">Allocation by the enterprise security architecture function</th> <th class="text-center" scope="col">Operational authority</th> </tr></thead><tbody><tr><td>CA-07 Continuous monitoring</td> <td>Implement an automated monitoring infrastructure</td> <td>Cyber security authorities (as a component of the cyber security function)</td> </tr><tr><td>SC-07 Boundary protection</td> <td>Implement a boundary protection infrastructure</td> <td>IT/OT operations group – Network operations</td> </tr><tr><td>CP-07 Alternate processing site</td> <td>Service contract</td> <td>Service provider</td> </tr><tr><td>IA-02 End user identification and authentication</td> <td>Subscribe to an existing <abbr title="Government of Canada">GC</abbr> shared identification and authentication service</td> <td>Shared Services Canada (SSC)<sup id="fn4-rf"><a class="fn-lnk" href="#fn4"><span class="wb-inv">Footnote </span>4</a></sup></td> </tr><tr><td>PT-01 Personal information handling and transparency policy and procedures</td> <td>The enterprise policy authority publishes policies and procedures for handling personal information</td> <td>Privacy authorities (as a component of the privacy function)</td> </tr></tbody></table></div> <p>When the allocation of a common security or privacy control or activity requires the implementation of supporting systems, organizations should implement them through projects following the process described in System lifecycle cyber security and privacy risk management activities (ITSP.10.037).</p> <p><strong>Output:</strong> Document listing the allocated common security and privacy controls and activities.</p> <h4 id="4.3.2">4.3.2 Allocate system-specific controls and activities</h4> <p><strong>Objective:</strong> Coordinate and document the allocation of system-specific controls and activities.</p> <p><strong>Suggested related roles:</strong> Enterprise security architects, privacy architects</p> <p>To implement the selected security and privacy controls and activities, the enterprise security architecture function will allocate a set of appropriate controls and activities to that system. System-specific technical controls must be identified early in the project lifecycle so they can be appropriately scoped, resourced and funded within the project budget. This proactive approach ensures that controls and activities are integrated seamlessly into system development and operations, rather than being retrofitted later at greater cost or complexity.</p> <p>The allocation process includes mapping applicable controls, including hybrid ones, to system contexts and coordinating implementation responsibilities across cyber security, privacy and business stakeholders.</p> <p>Projects and <abbr title="information technology and operational technology">IT/OT</abbr> operations groups implement and operate security and privacy controls and activities in systems following the cyber security and privacy risk management activities described in ITSP.10.037.</p> <p>Clear documentation of control and activity assignments also supports ongoing assurance, audit readiness, and the ability to manage the effectiveness of controls over time.</p> <p><strong>Output:</strong> A list of allocated security and privacy controls and activities in systems.</p> <!–** TOP OF PAGE ******–> <div class="clearfix"> </div> <div class="pull-right small text-muted mrgn-bttm-0"><a href="#wb-tphp">Top of page</a> <span aria-hidden="true" class="text-primary glyphicon glyphicon-circle-arrow-up"></span></div> <!–** END TOP OF PAGE **–> <h3 id="4.4">4.4 Monitor and assess performance of security and privacy controls and activities</h3> <p><strong>Objective:</strong> Perform continuous monitoring and assessment of the implemented security and privacy controls and activities.</p> <p><strong>Suggested related roles:</strong> <abbr title="information technology and operational technology">IT/OT</abbr> operations groups, security assessors, privacy assessors</p> <p>Organizations monitor and assess the performance of the implemented common, hybrid, and system-specific controls and activities through the collection, consolidation, and continuous analysis of performance metrics. Many of the monitoring and assessment tasks will be performed by the IT/OT operations groups, and summary reports will be provided to the cyber security and privacy functions for organization-wide analysis. Some sensitive monitoring tasks may be performed by the cyber security and privacy functions (if mandated by a security and privacy control and activity profile) when separation of duties is required.</p> <p>Continuous assessment<sup id="fn5-rf"><a class="fn-lnk" href="#fn5"><span class="wb-inv">Footnote </span>5</a></sup> tasks consider more than just the performance of security and privacy controls and activities and should include:</p> <ul><li>identifying vulnerabilities, weaknesses, and inefficiencies</li> <li>reviewing security and privacy-related changes</li> <li>reviewing security and privacy incidents and problems reports</li> <li>conducting security testing, which could include functional security testing, vulnerability assessments, and penetration testing</li> <li>reviewing the security configuration of system components</li> </ul><p>The monitoring and assessment results will inform organizational authorities of the current state of organizational systems’ security and privacy posture. A stable and adequate security and privacy posture will allow a system to maintain its authorization to operate. A deteriorating posture may lead organizational authorities to revoke the authorization to operate until mitigating measures are put in place (see <a href="#4.5">section 4.5</a> and <a href="#4.6">section 4.6</a>).</p> <p><strong>Output:</strong> Report on the performance of security and privacy controls and activities and on the security and privacy posture of the systems.</p> <h3 id="4.5">4.5 Maintain authorization</h3> <p><strong>Objective:</strong> Maintain the authorization state of the systems.</p> <p><strong>Suggested related roles:</strong> Authorizers, business analysts, security advisors, privacy advisors, enterprise security architects, privacy architects</p> <p>When following the suggested cyber security and privacy risk management process, organizations initially authorize the operation of their systems as part of the process described in ITSP.10.037. Subsequently, organizations maintain the authorization state of their systems by continuously monitoring, assessing, and updating security and privacy controls and activities.</p> <p>The authorization maintenance process consists of the following activities:</p> <ul><li>periodically reviewing the injury assessment and categorization of supported business activities</li> <li>reassessing the threat environment and the security performance of technical environments</li> <li>reassessing collection and use of personal information to understand any changes to the initial program design</li> <li>reviewing the performance assessment result for security and privacy controls and activities</li> <li>reviewing the activities of <abbr title="information technology and operational technology">IT/OT</abbr> operations group to ensure that they have adequately maintained the security and privacy posture of their systems according to the security and privacy provisions of their operations plans</li> </ul><p>When the results of authorization maintenance activities show that a system is no longer operating within acceptable levels of residual risk, there are several avenues that organizations can pursue to remediate the situation, including:</p> <ul><li>implementing temporary security measures to protect supported business activities (for example, disconnecting an information system from the Internet, activating a portion of a contingency plan)</li> <li>updating privacy artefacts (for example, privacy notice statements and associated personal information banks, subject to required approvals)</li> <li>updating implemented controls and activities to correct security and privacy deficiencies and return the system to its authorized state of operation (see <a href="#4.6">section 4.6</a>)</li> <li>accepting the new level of residual risk</li> </ul><p>If the level of residual risk remains unacceptable after initial remedial action, authorizers may choose to revoke the authority to operate pending further remedial action. The revocation of authorization would lead to additional security and privacy analysis activities to identify specific deficiencies within the operational context, followed by the application of corrective measures or improvements to implemented controls and activities to return the system to its authorized state.</p> <p>Organizations should establish in their organizational security plan the frequency of periodic security assessment and review activities (for example, review of security incident reports, review of threat environment, review of <abbr title="information technology and operational technology">IT/OT</abbr> operations group security activities). Alternatively, authorizers could establish the frequency of such activities during a system’s initial implementation according to factors such as security injury categorization, threats, residual risk levels, and outstanding security deficiencies.</p> <p>The outputs of authorization maintenance activities include updated residual risk assessments and updated security provisions in an operations plan. The security provisions of the operations plan should include mitigation plans and schedules for any outstanding security deficiencies discovered because of the security assessment work (see <a href="#4.6">section 4.6</a>).</p> <p>In the <abbr title="Government of Canada">GC</abbr>, there is an obligation for programs to annually review any risks identified during the privacy assessment process and update risk mitigation measures.</p> <p><strong>Outputs:</strong> Updated residual risk assessments and updated security provisions of operations plan.</p> <h3 id="4.6">4.6 Update security and privacy controls and activities profile(s)</h3> <p><strong>Objective:</strong> Ensure that the security and privacy posture of the program, including the system, remains adequate by keeping all implemented controls and activities up to date. If required, add controls and activities to increase the security and privacy posture.</p> <p><strong>Suggested related roles:</strong> Enterprise security architects, privacy architects</p> <p>Organizations may need to update their implemented security and privacy controls and activities for various reasons, including when there is a:</p> <ul><li>change in organizational missions or objectives</li> <li>change in an organizational business activity (for example, collection of new, more sensitive information under an existing organizational program)</li> <li>change in business needs for security and privacy (for example, because of legislative or policy changes)</li> <li>change in the collection, use or disclosure of personal information</li> <li>requirement for change because of an organizational threat assessment update (that is, when a program is targeted by more sophisticated threat actors)</li> <li>requirement for change because of performance monitoring (for example, a security or privacy control or activity has proven ineffective in adequately protecting related systems or personal information)</li> </ul><p>Key inputs to this activity include existing organizational control and activity profiles, threat repositories, performance monitoring results, as well as assessment and authorization maintenance activities (see <a href="#4.4">section 4.4</a> and <a href="#4.5">section 4.5</a>). This activity can reveal gaps in existing controls and activities, prompting updates or the addition of new ones. It may also lead to revisiting earlier cyber security and privacy activities to redefine requirements, update control and activity profiles and threat repositories (see, for example, <a href="#4.1.3">section 4.1.3</a>, <a href="#4.1.6">section 4.1.6</a>, <a href="#4.2">section 4.2</a>), allocate new controls or activities (see <a href="#4.3">section 4.3</a>), or conduct further monitoring and assessment (see <a href="#4.4">section 4.4</a>).</p> <p><strong>Output:</strong> Report on the need for improved or new security and privacy controls and activities or updated profile(s).</p> </section><!–** TOP OF PAGE ******–><div class="clearfix"> </div> <div class="pull-right small text-muted mrgn-bttm-0"><a href="#wb-tphp">Top of page</a> <span aria-hidden="true" class="text-primary glyphicon glyphicon-circle-arrow-up"></span></div> <!–** END TOP OF PAGE **–> <section><h2 class="text-info" id="5">5 Injury assessment and categorization process</h2> <p>This section describes how organizations can perform injury assessments and categorization of their organizational business activities to help define their security and privacy control and activity profiles. Business activities refer to business processes and related information assets, including personal information.</p> <p>Injury assessment and categorization is an instrument to establish the relative importance of organizational business activities. At the organizational level, injury categories of business activities serve as input for conducting threat assessments and developing control and activity profiles. At the system level, injury categories of business activities serve as input for establishing security assurance requirements, selecting and tailoring security and privacy controls and activities, and conducting <abbr title="threat risk assessment">TRA</abbr> activities.</p> <h3 id="5.1">5.1 Concepts</h3> <p>When business activities are compromised by cyber-related threats, there may be injury to the national interests<sup id="fn6-rf"><a class="fn-lnk" href="#fn6"><span class="wb-inv">Footnote </span>6</a></sup> or the non-national interests<sup id="fn7-rf"><a class="fn-lnk" href="#fn7"><span class="wb-inv">Footnote </span>7</a></sup> that the business activities serve. Injury can occur because of the unauthorized use, disclosure, modification, or denial of information, or the corruption or interruption of business processes. Injury due to the unauthorized use or disclosure of information relates to the <strong>confidentiality objective</strong> of information security. Injury due to the modification of information or the corruption of business processes relates to the <strong>integrity objective</strong> (that is, accuracy, completeness, authenticity, intended use) of information security. Injury due to the denial of information or the interruption of business processes relates to the <strong>availability objective</strong> of information security.</p> <p>The injury assessment and categorization process determines an injury level for each security objective associated with a business activity. The designation of an injury level assumes a successful compromise by a threat actor which allows the assessment to focus on the injuries that could reasonably result from this event.</p> <h4>5.1.1 National security systems</h4> <p>A national security system (NSS) is any system whose compromise, whether in terms of confidentiality, integrity or availability, would result in harm to national security. While <abbr title="national security system">NSS</abbr> designation often aligns with classification levels (such as confidential, secret, top secret), classification alone is not determinative. Some systems may be unclassified yet still qualify as <abbr title="national security system">NSS</abbr> if their compromise could reasonably cause injury to national security.</p> <p>In Canada, <abbr title="Treasury Board Secretariat">TBS</abbr> defines unclassified as not being sensitive. However, a system’s classification status does not negate its potential impact. For example, a military weapon system onboard a naval vessel or armoured vehicle, if compromised, could result in a national security injury. As such, it should be identified as an <abbr title="national security system">NSS</abbr> during the injury categorization process. The key consideration in <abbr title="national security system">NSS</abbr> designation is the potential for national security injury, not classification status. Injury assessments should prioritize the consequences of compromise over whether a system is labelled as classified or unclassified.</p> <h4>5.1.2 Safety</h4> <p>When assessing safety-critical systems, where failure could lead to outcomes such as health impacts, loss of life, property damage or loss, or environmental harm, it is essential to consider both the classification of information and how safety requirements influence the potential for harm in the event of system compromise. In such cases, safety requirements can elevate the potential impact on availability, integrity and, in some contexts, confidentiality, resulting in higher injury levels during the injury categorization process. Safety-critical systems usually have stringent expectations for availability and integrity, and confidentiality may also be important depending on the operational context. Ultimately, the severity of injury assigned should be commensurate with the level of harm that could realistically result from a compromise of the system.</p> <h4>5.1.3 Privacy</h4> <p>Injury assessments must account for the potential impact on individual privacy when business activities involve personal information. Compromised personal information can cause injury to individuals and organizations and can result in:</p> <ul><li>identity theft or fraud</li> <li>reputational or emotional harm</li> <li>loss of public trust</li> <li>legal and regulatory consequences</li> </ul><p>The level of injury depends on the sensitivity, volume, and context of the personal information. In addition to breaches of personal information, privacy injuries may also arise from corrupted personal data leading to incorrect decisions or disruption of services affecting access to benefits or critical support. This is especially significant in safety-critical or service-critical systems (for example, health or emergency services). Privacy risks should be considered alongside confidentiality, integrity, and availability objectives during the injury categorization process. Where personal information is used to make administrative decisions, a <abbr title="privacy impact assessment">PIA</abbr> should be conducted in parallel with threat and risk assessments to ensure appropriate privacy and security controls and activities are selected.</p> <h4>5.1.4 Adversarial perspective</h4> <p>Injury assessment and categorization require framing the compromise from an adversarial perspective. Security and privacy practitioners should evaluate injuries under the assumption that a threat actor has successfully compromised the system, including scenarios where they obtained significant control. Most organizations tend to underestimate potential injuries because they approach risk with good faith and reasonable assumptions. However, some threat actors are willing to engage in extraordinary and unconscionable actions to achieve their objectives.</p> <p>For example, a business process such as operating life support systems in a hospital should be assumed to potentially negatively impact patient health, including causing bodily harm and loss of life. Another example would be a department with a security mandate which, if compromised, could be leveraged as a criminal or hostile tool controlled by a threat actor, in complete opposition to their mandate.</p> <p>The question here is: What is the worst case-scenario?</p> <h4>5.1.5 Post-compromise injury</h4> <p>The injury assessment must be based on the assumption that an attack has occurred and has been successful. At this stage, the likelihood of the threat and the presence of mitigating controls should not be considered. When evaluating potential injury, even scenarios deemed exceedingly unlikely must be treated as if they have already occurred.</p> <p>For example, while the bombing of a data centre may be considered highly unlikely, the resulting injury – complete loss of all onsite functionality – remains a valid consideration in the assessment. Similarly, the existence of redundant data centres does not reduce the severity of the injury at this stage of the process.</p> <p>The question here is: What happens if that worst-case scenario occurred?</p> <!–** TOP OF PAGE ******–> <div class="clearfix"> </div> <div class="pull-right small text-muted mrgn-bttm-0"><a href="#wb-tphp">Top of page</a> <span aria-hidden="true" class="text-primary glyphicon glyphicon-circle-arrow-up"></span></div> <!–** END TOP OF PAGE **–> <h3 id="5.2">5.2 Injury assessment and categorization process</h3> <p>As illustrated in <a href="#fig3">figure 3</a>, the injury assessment and categorization process for a business activity can be summarized in 5 steps:</p> <ol><li>Develop a custom injury assessment table</li> <li>Identify the business processes and information assets, including personal information, that relate to the business activity</li> <li>Assess expected injuries from assumed threat compromise of the business activity, and determine their severity in terms of confidentiality, integrity, and availability</li> <li>Determine the injury category of the business activity by identifying the highest expected level injury across confidentiality, integrity, and availability</li> <li>Prepare an injury assessment and categorization report</li> </ol><!– IMAGE EN –><div class="panel panel-default"> <div class="panel-body"> <figure id="fig3"><figcaption class="mrgn-bttm-md"><strong>Figure 3 – Injury assessment and categorization process</strong></figcaption><img alt="Flow chart demonstrating injury assessment and categorization process – Long description immediately follows" class="img-responsive center-block mrgn-bttm-lg" src="/sites/default/files/images/itsp10036-fig3-e.png" /><details class="brdr-tp brdr-rght brdr-bttm brdr-lft mrgn-bttm-sm"><summary>Long description – Figure 3: Injury assessment and categorization process</summary><p>This figure represents a vertical flow chart explaining the injury assessment and categorization process. A single central column shows the five numbered steps of the process from top to bottom, each in a rectangle, connected by downward-pointing arrows. The process begins at an oval labeled "Start" and ends at an oval labeled "End." At several steps, inputs appear on the left side of the flow with arrows feeding into the step, and outputs appear on the right side as white document-shaped callouts. Each output callout has a small, numbered circle that serves as a persistent reference label used later in the diagram to show where that output is consumed as an input.</p> <p>The central sequence from top to bottom is:</p> <ul class="list-unstyled"><li>Start: An oval labeled "Start," with a downward arrow to Step 1.</li> <li>Step 1: Develop organization-specific injury assessment table <ul><li>Left input: A rectangular label "Strategic business documents" points to Step 1.</li> <li>Right output: A white document-shaped callout labeled "Organization-specific injury assessment table," marked with a small circle labeled "1." This creates Output 1.</li> <li>Note on reuse: Output 1 is later reused as an input to Step 3, indicated by a small circle "1" shown on the left side feeding into Step 3.</li> </ul></li> <li>Step 2: Identify business processes and related information assets <ul><li>Left input: A rectangular label "Documentation on business activities" points to Step 2.</li> <li>Right output: A white document-shaped callout labeled "Business processes and related information," marked with a small circle "2." This creates Output 2.</li> <li>Note on reuse: Output 2 is later reused as an input to Step 3 and Step 4, indicated by the small circle "2" on the left side of those steps.</li> </ul></li> <li>Step 3: Assess injuries from threat compromises <ul><li>Left inputs: <ul><li>Two rectangular sources feed into Step 3: "Impact assessments" and "Regulatory instruments."</li> <li>In addition, two small numbered circles also feed into Step 3 from the left margin: circle "1" (Output 1 from Step 1) and circle "2" (Output 2 from Step 2). These icons visually indicate that the organization-specific assessment table and the identified business processes/information assets are used within the injury assessment.</li> </ul></li> <li>Right outputs: <ul><li>A white callout labeled "Expected injuries from threat compromise," marked with circle "3." This creates Output 3.</li> <li>A second white callout labeled "Level of injuries as they relate to <abbr title="confidentiality, integrity and availability">CIA</abbr>," marked with circle "4." This creates Output 4.</li> </ul></li> <li>Note on reuse: Outputs 3 and 4 are used as inputs to Step 4, indicated by small circles "3" and "4" on the left side of Step 4.</li> </ul></li> <li>Step 4: Determine injury category of each business activity <ul><li>Left inputs: <ul><li>One vertical stack shows circles "2," "3," and "4," to indicate the reuse of Outputs 2, 3, and 4, to determine the injury category.</li> </ul></li> <li>Right output: <ul><li>A white callout labeled "High watermark injury category of each business activity," marked with circle "5." This creates Output 5. "High watermark" conveys that the injury category is set to the highest expected severity across confidentiality, integrity, and availability.</li> </ul></li> <li>Note on reuse: Output 5 is used as an input to Step 5, indicated by circle "5."</li> </ul></li> <li>Step 5: Prepare injury categorization report <ul><li>Left inputs: A vertical sequence of small circles "1, 2, 3, 4, 5" shows that the final report compiles all earlier outputs: Output 1 (organization-specific table), Output 2 (business processes/information), Output 3 (expected injuries), Output 4 (<abbr title="confidentiality, integrity and availability">CIA</abbr>-level severity), and Output 5 (high watermark injury category).</li> <li>No right-side callout is drawn; the flow continues to the end.</li> </ul></li> <li>End: An oval labeled "End" below Step 5 marks the conclusion of the process.</li> </ul></details></figure></div> </div> <!– END IMAGE EN –> <div class="panel panel-info"> <header class="panel-heading"><h3 class="panel-title"><abbr title="Government of Canada">GC</abbr>-specific note</h3> </header><div class="panel-body"> <p>The goal of the security categorization process is to determine the sensitivity and criticality of business activities based on an injury test. This is similar to the goal of the sensitivity assessment questionnaire that is found in the statement of sensitivity (SoS) of many departments and agencies. An <abbr title="statement of sensitivity">SoS</abbr> captures the types of information assets that an information system will store and process and the classification of that information (for example, Protected B). The <abbr title="statement of sensitivity">SoS</abbr> also collects the general importance of integrity, as well as the criticality of the program or services, which is often expressed as the maximum permissible downtime.</p> <p>Although injury is reflected in the information classification marking, the security categorization process makes the injury test for confidentiality explicit. Moreover, the security categorization process extends this explicit injury test to integrity and availability. The result is a security marking that reflects the expected level of injury for confidentiality, integrity, and availability (for example, Protected B, medium integrity, medium availability or <abbr title="Protected B, medium integrity, medium availability">PBMM</abbr>).</p> <p>In departments and agencies where the <abbr title="statement of sensitivity">SoS</abbr> process has been successfully implemented and has been in use for many years, completely abandoning a familiar process may not be the best approach to transition from the <abbr title="statement of sensitivity">SoS</abbr> process to the security categorization process. Departments and agencies could leverage familiarity and acceptability of their <abbr title="statement of sensitivity">SoS</abbr> process by simply incorporating an injury test in the confidentiality, integrity, and availability sections of their sensitivity assessment questionnaire.</p> </div> </div> <!–** TOP OF PAGE ******–> <div class="clearfix"> </div> <div class="pull-right small text-muted mrgn-bttm-0"><a href="#wb-tphp">Top of page</a> <span aria-hidden="true" class="text-primary glyphicon glyphicon-circle-arrow-up"></span></div> <!–** END TOP OF PAGE **–> <h3 id="5.3">5.3 Injury assessment and categorization process description</h3> <p>The organization should establish a dedicated working group to effectively carry out the injury assessment and categorization process. This group should typically include executives with operational responsibilities, senior security officials (such as the <abbr title="chief security officer">CSO</abbr> or their delegates), senior privacy officials (such as the <abbr title="chief privacy officer">CPO</abbr> or their delegates), security and privacy advisors, and business analysts. Depending on the organization and context, other roles could include a risk management lead, a legal counsel, a compliance officer, security and privacy architects, and a communications or public relations lead.</p> <h4>5.3.1 Develop organization-specific injury assessment table</h4> <p>To conduct injury assessments and categorization, organizations should create their own injury tables. Small organizations with 1 or few programs serving common interests may only need 1 injury table while large organizations with several programs serving different interests may require several injury tables. If financial thresholds are incorporated into the tailored injury table, the specified dollar amounts should be proportionate to the organization’s overall budget to ensure contextual relevance. For example, in the <abbr title="Government of Canada">GC</abbr>, Public Services and Procurement Canada (PSPC)’s Federal Pension Administration program serves very different interests than PSPC’s Linguistic Services.</p> <p>To develop an organization-specific injury assessment table that will inform future risk assessments, it is recommended to consult strategic business documents, such as corporate plans, performance commitments, and priority-setting frameworks, that articulate the organization’s objectives and obligations. Business analysts should be closely involved in this process, as they possess critical contextual knowledge and can help identify realistic and relevant injury scenarios. Additional sources, such as regulatory filings, stakeholder reports, and internal audit findings, may also provide valuable insight into business impacts and priorities.</p> <p><a href="#tab2">Table 2</a> contains examples of injury types and injury levels to help <abbr title="Government of Canada">GC</abbr> departments develop injury tables. Departments should include in their injury tables only those types and levels that relate to the interests served by their programs, services, and activities.</p> <div class="table-responsive"> <table class="table table-bordered"><caption class="text-center" id="tab2">Table 2: Examples of injury types and levels (<abbr title="Government of Canada">GC</abbr>-centric)</caption> <thead><tr class="active"><th class="text-center" scope="col">Injury type</th> <th class="text-center" scope="col">Very low injury level</th> <th class="text-center" scope="col">Low injury level</th> <th class="text-center" scope="col">Medium injury level</th> <th class="text-center" scope="col">High injury level</th> <th class="text-center" scope="col">Very high injury level</th> </tr></thead><tbody><tr><th scope="row"><strong>Civil disorder or unrest</strong></th> <td>No reasonable or negligible expectation of injury</td> <td>Civil disobedience, public obstructions</td> <td>Riot</td> <td>Sabotage of critical assets (for example, critical infrastructure)</td> <td>Large-scale civil unrest or sabotage requiring martial law</td> </tr><tr><th scope="row"><strong>Physical harm to people</strong></th> <td>No reasonable or negligible expectation of injury</td> <td>Physical discomfort</td> <td>Physical pain, injury, trauma, hardship, illness</td> <td>Physical disability, loss of life</td> <td>Widespread loss of life</td> </tr><tr><th scope="row"><strong>Psychological harm to people</strong></th> <td>No reasonable or negligible expectation of injury</td> <td>Stress</td> <td>Psychological distress</td> <td>Mental health disorder</td> <td>Widespread or severe mental health harm</td> </tr><tr><th scope="row"><strong>Financial loss to individuals</strong></th> <td>No reasonable or negligible expectation of injury</td> <td>Stress or discomfort</td> <td>Reduced quality of life</td> <td>Loss of financial security</td> <td>Not applicable</td> </tr><tr><th scope="row"><strong>Financial loss to Canadian companies</strong></th> <td>No reasonable or negligible expectation of injury</td> <td>Reduced performance</td> <td>Reduced competitiveness</td> <td>Loss of viability</td> <td>Not applicable</td> </tr><tr><th scope="row"><strong>Financial loss to the Canadian government</strong></th> <td>No reasonable<br /> or negligible expectation of injury</td> <td>Reduced program performance</td> <td>Degraded program outcomes</td> <td>Loss of program viability</td> <td>Loss of viability of key programs</td> </tr><tr><th scope="row"><strong>Harm to the Canadian economy</strong></th> <td>Not applicable</td> <td>Not applicable</td> <td>Reduced performance</td> <td>Reduced international competitiveness</td> <td>Compromise of key economic sectors</td> </tr><tr><th scope="row"><strong>Harm to Canada’s reputation</strong></th> <td>No reasonable or negligible expectation of injury</td> <td>Loss of Canadian public confidence</td> <td>Embarrassment (home or abroad)</td> <td>Damage to federal-provincial relations</td> <td>Damage to diplomatic or international relations</td> </tr><tr><th scope="row"><strong>Loss of Canadian sovereignty</strong></th> <td>Not applicable</td> <td>Not applicable</td> <td>Impediments to the development of major government policies</td> <td>Impediments to effective law enforcement<br /> Loss of continuity of government</td> <td>Loss of territorial sovereignty</td> </tr></tbody></table></div> <p><a href="#tab3">Table 3</a> presents examples of injury types and injury levels that are more generally applicable. Organizations should include in their injury tables only those types and levels that relate to the interests served by their programs, services, and activities.</p> <div class="table-responsive"> <table class="table table-bordered"><caption class="text-center" id="tab3">Table 3: Examples of injury types and levels</caption> <thead><tr class="active"><th class="text-center" scope="col">Injury type</th> <th class="text-center" scope="col">Very low injury level</th> <th class="text-center" scope="col">Low injury level</th> <th class="text-center" scope="col">Medium injury level</th> <th class="text-center" scope="col">High injury level</th> <th class="text-center" scope="col">Very high injury level</th> </tr></thead><tbody><tr><th scope="row"><strong>Protest against the organization</strong></th> <td>No reasonable or negligible expectation of injury</td> <td>Civil disobedience, public obstructions</td> <td>Riot</td> <td>Destruction of property assets</td> <td>Not applicable</td> </tr><tr><th scope="row"><strong>Physical harm to people</strong></th> <td>No reasonable or negligible expectation of injury</td> <td>Physical discomfort</td> <td>Physical pain, injury, trauma, hardship, illness</td> <td>Physical disability, loss of life</td> <td>Widespread loss of life</td> </tr><tr><th scope="row"><strong>Psychological harm to people</strong></th> <td>No reasonable or negligible expectation of injury</td> <td>Stress</td> <td>Psychological distress</td> <td>Mental health disorder</td> <td>Widespread or severe mental health harm</td> </tr><tr><th scope="row"><strong>Financial loss to individuals</strong></th> <td>No reasonable or negligible expectation of injury</td> <td>Stress or discomfort</td> <td>Reduced quality of life</td> <td>Loss of financial security</td> <td>Not applicable</td> </tr><tr><th scope="row"><strong>Financial loss to the organization</strong></th> <td>No reasonable<br /> or negligible expectation of injury</td> <td>Reduced performance</td> <td>Reduced competitiveness</td> <td>Loss of viability</td> <td>Bankruptcy</td> </tr><tr><th scope="row"><strong>Legal/regulatory actions against the organization</strong></th> <td>No legal or regulatory actions expected<br /> Warning notices</td> <td>Small fines<br /> Small redress expenses</td> <td>Moderate fines<br /> Stop work orders<br /> Moderate redress expenses</td> <td>Significant fines<br /> Temporary licence revocation<br /> Limitations on activities<br /> Significant redress expenses</td> <td>Severe fines<br /> Permanent licence revocation<br /> Cessation order</td> </tr><tr><th scope="row"><strong>Civil actions</strong></th> <td>No reasonable or negligible expectation of liability<br /> Settlement with client for a nominal amount</td> <td>Minimal liability<br /> Out-of-court settlement</td> <td>Moderate liability<br /> Compensation</td> <td>Significant liability<br /> High-profile lawsuit<br /> Significant financial settlements</td> <td>Severe liability<br /> Class-action lawsuit<br /> Severe financial penalties<br /> Loss of clients’ trust</td> </tr><tr><th scope="row"><strong>Reputational harm</strong></th> <td>No reasonable or negligible expectation of injury</td> <td>Minimal reputational impact<br /> Minimal reputational management costs<br /> Limited negative media or social media coverage</td> <td>Moderate reputational impact requiring corrective action<br /> Increased costs to maintain and attract clients<br /> Sustained negative media or social media coverage</td> <td>Significant reputational harm requiring immediate corrective action<br /> Significant costs to maintain and attract clients<br /> Widespread or prolonged negative media or social media coverage<br /> Loss of clients expected</td> <td>Severe and potentially irreparable reputational harm<br /> Loss of key clients<br /> Long-term decline in brand reputation</td> </tr></tbody></table></div> <h4>5.3.2 Identify business processes and related information assets</h4> <p>The second step of the injury assessment and categorization process is to identify the business processes and related information assets, including personal information, relevant to the business activity.</p> <p>There are several sources from which to identify and describe business processes and related information assets. Good sources include:</p> <ul><li>business cases</li> <li><abbr title="concept of operations">CONOPS</abbr></li> <li>functional business requirements</li> <li>enterprise architecture documentation which typically describes an organization’s business processes and related information assets</li> <li>discussions or interviews with business analysts and other individuals within related business communities</li> </ul><p>The output of this step is a brief description of the business processes and related information assets of relevance to the business activity.</p> <h4>5.3.3 Assess injuries from threat compromise</h4> <p>The third step of the process is the injury assessment. The objective is to determine the potential injuries from threat compromises for each of the business processes and related information assets identified in the previous step. This is achieved by first determining the injuries that could possibly occur as a result of threat actors compromising the confidentiality, integrity, and availability of the business processes and related information assets, and then attributing appropriate levels to these injuries.</p> <p>Generally, confidentiality, integrity, and availability apply to information assets, while integrity and availability apply to business processes. In the <abbr title="Government of Canada">GC</abbr>, in some exceptional cases where national security or national defence is involved, confidentiality can apply to business processes. In the private sector, those exceptions would usually be related to trade secrets.</p> <p>Injury levels that apply to business activities may already be documented in organizational deliverables such as business impact analyses, <abbr title="privacy impact assessments">PIAs</abbr>, <abbr title="statement of sensitivity">SoS</abbr>, business risk assessments, and <abbr title="threat risk assessments">TRAs</abbr>. It may also be helpful to review regulatory instruments such as laws, policies, and regulations that may apply to specific business processes or information. Non-compliance with regulatory requirements could, under certain circumstances, lead to penalties or sanctions that could increase injury. Organizations can leverage this information when assessing injuries related to a failure of their business activities.</p> <p>When assessing injuries, security practitioners should consider several factors that may influence the results, including aggregation, inference, and interdependency.</p> <h5>5.3.3.1 Aggregation</h5> <p>As the number of assets increases, the injuries arising from compromise may also grow. For example, the unauthorized disclosure of a single personnel file might be expected to cause some embarrassment to the individual and generate public anxiety regarding an organization’s ability to protect personal information. However, if all human resources records for a major organization were released inappropriately, the adverse effects could be significantly worse.</p> <p>From a confidentiality perspective, aggregation has 2 dimensions:</p> <ul><li>increased sensitivity as more data elements are added to a record</li> <li>increased sensitivity as more records are collected in a file or database</li> </ul><p>In each case, the confidentiality value of the whole may be greater than that of the individual parts, based on the increased injury expected in the event of unauthorized disclosure.</p> <p>Aggregation applies equally to availability and integrity values. For example, the destruction of 1 asset, such as a single vehicle, might have clearly defined consequences, whereas the loss of an entire fleet would be much more serious. Unauthorized modification of a single record or complete corruption of a large database would be the integrity equivalent.</p> <h5>5.3.3.2 Inference</h5> <p>By its very nature, inference applies only to confidentiality, where access to seemingly innocuous data of little or no sensitivity may allow a knowledgeable individual to draw much more damaging conclusions regarding another organization’s capabilities or intentions. For example, studying personnel movements to identify the qualifications and affiliations of executives and other employees selected for different positions may provide useful indicators of future policy platforms or business plans.</p> <p>Unfortunately, the relationships among various data and data sources may be very subtle, even tenuous, so opportunities for inference are generally far more difficult to identify and assess. From a practical point of view, inference must be assessed cautiously because complex scenarios can exaggerate the risk and lead to needless overclassification. The inference scenarios of concern are the ones where aggregations in a non-national security system lead to national security injuries.</p> <h5>5.3.3.3 Interdependency</h5> <p>Due to interdependencies, the loss or degradation of 1 business process and its associated information may impact other processes and related information. The purpose of analyzing interdependencies is to determine if there is a likelihood of a high cascading effect on other processes and information resulting from the compromise of a business process or information.</p> <p>Similar to the problem of aggregation, the injury that would result from the cascading loss of 1 element may be greater than the injury level assigned to any of the independent elements. Types of interdependencies include:</p> <ul><li>physical (for example, material output of an infrastructure used by another)</li> <li>geographic (for example, a common corridor)</li> <li>logical (for example, dependency through financial markets)</li> </ul><p>The output of this step is a list of expected injuries and injury levels for confidentiality, integrity, and availability by business process and related information assets.</p> <div class="panel panel-info"> <header class="panel-heading"><h3 class="panel-title"><abbr title="Government of Canada">GC</abbr>-specific note</h3> </header><div class="panel-body"> <p>For consistency within and across departments, security and privacy practitioners should adopt a common categorization marking scheme. As a guideline, it is recommended that injury categories be expressed using the following marking format:</p> <ul><li>Confidentiality: Protected/Classified Level</li> <li>Integrity: Very low/Low/Medium/High/Very high</li> <li>Availability: Very low/Low/Medium/High/Very high</li> </ul><p>For example: Protected B, Medium, Medium (PBMM)</p> </div> </div> <h4>5.3.4 Determine injury category of each business activity</h4> <p>The fourth step in the injury assessment and categorization process is to determine the injury category of each business activity.</p> <p>In normal circumstances, the injury category of a business activity should express the highest levels of injury of all related business processes and information assets for each of the security objectives. Individually, these elements may be attributed different levels of injury for a given protection objective. For example, a business activity may involve 1 type of information with an assessed injury level of low for confidentiality and another type of information with an assessed injury level of medium for the same security objective (both for non-national interest). These individual values are important and should be documented. However, the injury category of the business activity should reflect the highest level of injury. For the preceding example, the business activity’s confidentiality level would be marked as "Medium".</p> <p>Notwithstanding, there may be circumstances where more analysis is required to determine the most appropriate injury category. For example, security and privacy practitioners may attribute a higher level than the "High" benchmark because of the aggregate effects of threat compromise, or an interdependency involving a critical process outside of a business activity’s boundary.</p> <p>The output of this step is the injury category of each business activity, which can be expressed using the same categorization marking format as for individual business processes and information.</p> <h4>5.3.5 Prepare injury categorization report</h4> <p>The final step of the injury assessment and categorization process is to prepare the injury categorization report.</p> <p>Security and privacy practitioners should summarize in a report the results of the injury assessment for reporting purposes and to serve as input to the organizational security and privacy control and activity profile development. For each business process and related information, the injury categorization report should include:</p> <ul><li>a short description of the business process and related information asset(s)</li> <li>a description of the expected injuries from threat compromise</li> <li>the levels of expected injury as they relate to the loss of confidentiality, integrity, and availability</li> <li>the rationale for attributing the levels of injury</li> </ul><h3 id="5.4">5.4 Examples</h3> <p>A spreadsheet with simple examples that help illustrate the injury assessment and categorization process for business activities is available. To request a copy of this spreadsheet, <a href="/en/contact-cyber-centre">contact the Contact Centre</a>.</p> </section><!–** TOP OF PAGE ******–><div class="clearfix"> </div> <div class="pull-right small text-muted mrgn-bttm-0"><a href="#wb-tphp">Top of page</a> <span aria-hidden="true" class="text-primary glyphicon glyphicon-circle-arrow-up"></span></div> <!–** END TOP OF PAGE **–> <h2 class="text-info" id="ann.a">Government of Canada injury levels conversion</h2> <p>Other injury assessment frameworks may use a "3-level" scale of Low, Medium, and High instead of the 5-level scale presented in this publication. Where this is the case, departments can use the scale conversions specified in <a href="#tab4">Table 4</a>, <a href="#tab5">Table 5</a> and <a href="#tab6">Table 6</a> to convert injury levels from their 3- or 4-level scale to the 5-level scale outlined in ITSP.10.036.</p> <p>In conversion <a href="#tab6">Table 6</a>, the "Low" category from the 3-level injury scale may correspond to either "Very low" or "Low" on the 5-level scale, while "High" may map to either "High" or "Very high". To ensure appropriate categorization, it is recommended to re-evaluate the injury assessment using the 5-level scale directly, considering the specific context and desired level of granularity.</p> <div class="table-responsive"> <table class="table table-bordered"><caption class="text-center" id="tab4">Table 4: Confidentiality – Non-national interest (Protected)</caption> <thead><tr class="active"><th class="text-center" scope="col">4-level scale value</th> <th class="text-center" scope="col">5-level scale value</th> </tr></thead><tbody><tr><td>No expected injury – Unclassified</td> <td>Very low</td> </tr><tr><td>Low – Protected A</td> <td>Low</td> </tr><tr><td>Medium – Protected B</td> <td>Medium</td> </tr><tr><td>High – Protected C</td> <td>High</td> </tr><tr><td>Not applicable</td> <td>Very high</td> </tr></tbody></table></div> <div class="clearfix"> </div> <div class="table-responsive"> <table class="table table-bordered"><caption class="text-center" id="tab5">Table 5: Confidentiality – National interest (Classified)</caption> <thead><tr class="active"><th class="text-center" scope="col">4-level scale value</th> <th class="text-center" scope="col">5-level scale value</th> </tr></thead><tbody><tr><td>No expected injury – Unclassified</td> <td>Very low</td> </tr><tr><td>Not applicable</td> <td>Low</td> </tr><tr><td>Low – Confidential</td> <td>Medium</td> </tr><tr><td>Medium – Secret</td> <td>High</td> </tr><tr><td>High – Top Secret</td> <td>Very high</td> </tr></tbody></table></div> <div class="clearfix"> </div> <div class="table-responsive"> <table class="table table-bordered"><caption class="text-center" id="tab6">Table 6: Integrity and availability</caption> <thead><tr class="active"><th class="text-center" scope="col">4-level scale value</th> <th class="text-center" scope="col">5-level scale value</th> </tr></thead><tbody><tr><td>No injury expected<br /> Low</td> <td>Very low</td> </tr><tr><td>Low</td> <td>Low</td> </tr><tr><td>Medium</td> <td>Medium</td> </tr><tr><td>High</td> <td>High</td> </tr><tr><td>High</td> <td>Very high</td> </tr></tbody></table></div> <div class="clearfix"> </div> <!–FOOTNOTE SECTION EN–> <aside class="wb-fnote" role="note"><h2 id="reference">References</h2> <dl><dt>Footnote 1</dt> <dd id="fn1"> <p>Departments as defined in section 2 and entities included in Schedules IV and V of the <a href="https://laws.justice.gc.ca/eng/acts/F-11/index.html"><em>Financial Administration Act</em></a>, unless excluded by specific acts, regulations or orders in council.</p> <p class="fn-rtn"><a href="#fn1-rf"><span class="wb-inv">Return to footnote</span>1<span class="wb-inv"> referrer</span></a></p> </dd> <dt>Footnote 2</dt> <dd id="fn2"> <p>Personal information is defined in Section 3 of the <em>Privacy Act</em> as information about an identifiable individual, recorded in any manner.</p> <p class="fn-rtn"><a href="#fn2-rf"><span class="wb-inv">Return to footnote</span>2<span class="wb-inv"> referrer</span></a></p> </dd> <dt>Footnote 3</dt> <dd id="fn3"> <p>An organizational threat assessment is an ongoing process that also supports the system cyber security and privacy risk management activities. For more details, read System lifecycle cyber security and privacy risk management activities (ITSP.10.037).</p> <p class="fn-rtn"><a href="#fn3-rf"><span class="wb-inv">Return to footnote</span>3<span class="wb-inv"> referrer</span></a></p> </dd> <dt>Footnote 4</dt> <dd id="fn4"> <p>Designated provider of <abbr title="Government of Canada">GC</abbr> shared services.</p> <p class="fn-rtn"><a href="#fn4-rf"><span class="wb-inv">Return to footnote</span>4<span class="wb-inv"> referrer</span></a></p> </dd> <dt>Footnote 5</dt> <dd id="fn5"> <p>Continuous assessment does not necessarily mean real-time assessment. Some assessment activities, such as log event review, can be performed automatically in real time, while other assessment activities, such as the review of system backup procedures, would usually be done manually on a pre-established schedule.</p> <p class="fn-rtn"><a href="#fn5-rf"><span class="wb-inv">Return to footnote</span>5<span class="wb-inv"> referrer</span></a></p> </dd> <dt>Footnote 6</dt> <dd id="fn6"> <p>National interests: The security and the social, political, and economic stability of Canada.</p> <p class="fn-rtn"><a href="#fn6-rf"><span class="wb-inv">Return to footnote</span>6<span class="wb-inv"> referrer</span></a></p> </dd> <dt>Footnote 7</dt> <dd id="fn7"> <p>Non-national interests: The safety, health, and well-being of individuals, and the financial position and reputation of individuals and Canadian companies.</p> <p class="fn-rtn"><a href="#fn7-rf"><span class="wb-inv">Return to footnote</span>7<span class="wb-inv"> referrer</span></a></p> </dd> </dl></aside></div> </div> </div> </div> </div> </article>

<article data-history-node-id="8017" about="/en/guidance/careful-adoption-agentic-ai" class="cccs-basic-page full clearfix"> <div class="content"> <div class="layout layout–onecol"> <div class="layout__region layout__region–content"> <div data-block-plugin-id="extra_field_block:node:cccs_basic_page:links" class="block block-layout-builder block-extra-field-blocknodecccs-basic-pagelinks clearfix"> </div> <div data-block-plugin-id="field_block:node:cccs_basic_page:body" class="block block-layout-builder block-field-blocknodecccs-basic-pagebody clearfix"> <div class="field field–name-body field–type-text-with-summary field–label-hidden field–item"><section><p>Agentic artificial intelligence (AI) systems increasingly operate across critical infrastructure and defence sectors and support mission-critical systems. As agentic <abbr title="artificial intelligence">AI</abbr> systems play a growing operational role, it is crucial for defenders to implement security controls to protect national security and critical infrastructure systems from agentic <abbr title="artificial intelligence">AI</abbr>-specific risks.</p> <p>Agentic <abbr title="artificial intelligence">AI</abbr> can automate repetitive, well-defined and low-risk tasks. However, these additional opportunities come with additional risks. Like other <abbr title="artificial intelligence">AI</abbr> services, agentic <abbr title="artificial intelligence">AI</abbr> can be misused or misappropriated, leading to productivity losses, service disruption, privacy breaches or cyber security incidents. Organizations must therefore anticipate what could go wrong, assess how agentic <abbr title="artificial intelligence">AI</abbr> risk scenarios might affect operations and establish ongoing visibility and assurance to maintain confidence in their agentic <abbr title="artificial intelligence">AI</abbr> investments. Where possible, organizations should also consider a full spectrum of solutions for repetitive tasks, including reducing or eliminating low-value processes, which may be lower risk compared to agentic <abbr title="artificial intelligence">AI</abbr> solutions.</p> <p>The Canadian Centre for Cyber Security (Cyber Centre) joined the following international partners to co-author this publication:</p> <ul><li>Australian Signals Directorate’s Australian Cyber Security Centre (ASD’s ACSC)</li> <li>United States’ Cybersecurity and Infrastructure Security Agency (CISA)</li> <li>United States’ National Security Agency (NSA)</li> <li>New Zealand’s National Cyber Security Centre (NCSC-NZ)</li> <li>United Kingdom’s National Cyber Security Centre (NCSC-UK)</li> </ul><p>Throughout this publication, these organizations are referred to as the authoring agencies. The guidance below discusses key cyber security challenges and risks associated with the introduction of agentic <abbr title="artificial intelligence">AI</abbr> into information technology (IT) environments, as well as best practices for securing agentic <abbr title="artificial intelligence">AI</abbr> systems.</p> <p>The authoring agencies strongly recommend aligning agentic <abbr title="artificial intelligence">AI</abbr> risks and mitigation strategies with your organization’s existing security model and risk posture. The authoring agencies further recommend adopting agentic <abbr title="artificial intelligence">AI</abbr> cautiously by carefully assessing its use and never granting it broad or unrestricted access, especially to sensitive data or critical systems. Additionally, organizations should only use agentic <abbr title="artificial intelligence">AI</abbr> for low-risk and non-sensitive tasks.</p> </section><section><details class="mrgn-tp-md"><summary><h2 class="h3">Table of contents</h2> </summary><ul class="list-unstyled"><li><a href="#Scope">Scope and audience</a></li> <li><a href="#Agentic-ai">Agentic AI explained</a> <ul><li><a href="#fig1">Figure 1: Agentic AI system diagram</a></li> <li><a href="#Gen-agentic">Difference between agentic and generative AI</a></li> </ul></li> <li><a href="#Broader-considerations">Broader agentic AI security considerations</a> <ul><li><a href="#Inherited-risks">Inherited risks of large language models </a></li> <li><a href="#Increased-attack">Increased attack surface</a></li> <li><a href="#Complexity">Increased complexity</a></li> <li><a href="#Evolving-security">Evolving security as technology matures </a></li> <li><a href="#AI-security">AI security as part of cyber security</a></li> </ul></li> <li><a href="#Security-risks">Agentic AI security risks</a> <ul><li><a href="#Priv-risk">Privilege risks</a></li> <li><a href="#Design-config">Design and configuration risks</a></li> <li><a href="#Behav-risk">Behavioural risks</a></li> <li><a href="#Structural-risk">Structural risks</a></li> <li><a href="#Accountability-risk">Accountability risks</a></li> </ul></li> <li><a href="#Best-practices">Best practices for securing agentic AI systems</a></li> <li><a href="#conclusion">Conclusion</a></li> </ul></details></section><section><h2 id="Scope">Scope and audience</h2> <p>This publication primarily focuses on large language model (LLM)-based agentic <abbr title="artificial intelligence">AI</abbr> systems. It considers threats to and vulnerabilities within agentic <abbr title="artificial intelligence">AI</abbr> systems, as well as risks arising from agentic <abbr title="artificial intelligence">AI</abbr> behaviour. This includes risks introduced through system components, integrations and downstream use.</p> <p>The authoring agencies developed this publication to support government, critical infrastructure and industry stakeholders in understanding the key security challenges and risks posed by agentic <abbr title="artificial intelligence">AI</abbr>. It provides practical guidance to help organizations design, develop, deploy and operate agentic <abbr title="artificial intelligence">AI</abbr> systems and to make informed risk assessments and mitigations. The publication concludes with actionable recommendations to help organizations prepare for and defend against emerging and future agentic <abbr title="artificial intelligence">AI</abbr> threats.</p> </section><section><h2 id="Agentic-ai">Agentic AI explained</h2> <p>Agentic <abbr title="artificial intelligence">AI</abbr> systems are composed of one or more agents that fundamentally rely on an <abbr title="large language model">LLM</abbr> to interpret and reason about the state of the world, make decisions and take actions. Figure 1 shows how agentic <abbr title="artificial intelligence">AI</abbr> systems contain the <abbr title="large language model">LLM</abbr> itself, alongside external tools, external data sources, memory and planning workflows. These components enable the system to perceive its environment and, where applicable, take action to achieve its goals. Compared with traditional <abbr title="large language model">LLM</abbr> systems, agentic <abbr title="artificial intelligence">AI</abbr> systems distinguish themselves by accomplishing underspecified objectives, acting autonomously, following goal-directed behaviours and creating long-term plans.</p> <p>Agentic <abbr title="artificial intelligence">AI</abbr> systems are intended to operate without continuous human intervention. While a human typically designs and configures the system, some agentic <abbr title="artificial intelligence">AI</abbr> systems are also capable of autonomously creating, or spawning, sub-agents to accomplish specific sub-tasks.</p> <p>System design includes defining goals, providing conditions on which to act (called triggers) and making information available to the <abbr title="artificial intelligence">AI</abbr> service. Agents have some key attributes, including:</p> <ul><li>information input, such as user input, operating context and configuration parameters</li> <li>measurable goals identified from user directions, such as “minimize downtime for this server”</li> <li>statistical models, such as LLMs to identify what actions to take</li> <li>action and execution privileges, such as permissions to interact with tools, users, systems and operating environments</li> <li>tool or service access, such as system software and interfaces to take identified actions</li> <li>metrics, such as measurable indicators used by the designer to evaluate operational effectiveness and improve efficiency</li> </ul><div class="panel-body"> <h2 class="text-center h5" id="fig1">Figure 1. Agentic <abbr title="artificial intelligence">AI</abbr> system diagram</h2> <figure><img alt="Figure 1 – Long description immediately follows" class="img-responsive center-block mrgn-bttm-lg" src="/sites/default/files/images/agentic-ai-systems-1000×286.jpg" /></figure><details><summary>Long Description – Figure 1: Agentic AI system diagram</summary><p>Agentic <abbr title="artificial intelligence">AI</abbr> system diagram shows how agentic <abbr title="artificial intelligence">AI</abbr> systems contain the <abbr title="large language model">LLM</abbr> itself, alongside external tools, external data sources, memory and planning workflows. These components enable the system to perceive its environment and, where applicable, take action to achieve its goals. Compared with traditional <abbr title="large language model">LLM</abbr> systems, agentic <abbr title="artificial intelligence">AI</abbr> systems distinguish themselves by accomplishing underspecified objectives, acting autonomously, following goal-directed behaviours and creating long-term plans.</p> </details></div> <div> <h3 id="Gen-agentic">Difference between agentic and generative AI</h3> <p>Generative <abbr title="artificial intelligence">AI</abbr> (GenAI) is a subset of <abbr title="artificial intelligence">AI</abbr> that creates new content based on complex patterns learned from large datasets. <abbr title="Generative AI">GenAI</abbr> is commonly used to generate text, images, audio and video intended for human use or action. In contrast, agentic <abbr title="artificial intelligence">AI</abbr> builds on <abbr title="Generative AI">GenAI</abbr> by integrating with software systems to create autonomous agents that can independently reason, plan and take actions without requiring human intervention.</p> </div> </section><!–** TOP OF PAGE ******–><div class="pull-right small text-muted mrgn-bttm-0"><a href="#wb-tphp">Top of page</a> <span aria-hidden="true" class="text-primary glyphicon glyphicon-circle-arrow-up"></span></div> <!–** END TOP OF PAGE **–> <section><h2 id="Broader-considerations">Broader agentic AI security considerations</h2> <p>Organizations should review the following considerations when evaluating the broader concerns and risks associated with agentic <abbr title="artificial intelligence">AI</abbr>.</p> <div> <h3 id="Inherited-risks">Inherited risks of large language models</h3> <p>Since the core of agentic <abbr title="artificial intelligence">AI</abbr> is an LLM, agents inherit <abbr title="large language model">LLM</abbr> vulnerabilities. For example, actors could perform prompt injection attacks by including malicious prompts in phishing emails to convince email-monitoring agents to download malware. This underscores a key vulnerability: malicious actors can target agentic <abbr title="artificial intelligence">AI</abbr> systems using existing <abbr title="artificial intelligence">AI</abbr> and cyber attack vectors.</p> </div> <div> <h3 id="Increased-attack">Increased attack surface</h3> <p>Agentic <abbr title="artificial intelligence">AI</abbr> systems rely on a variety of components, including tools, external data sources and memory bases, to interact with their environment and expand their capabilities. Each of these components can introduce vulnerabilities across an interconnected attack surface that malicious actors can exploit. For instance, external data sources such as a web search can insert additional information into the prompt context, enabling indirect prompt injection attacks. With broader access to computing infrastructure, malicious actors may exploit system components to conduct attacks, such as executing malicious scripts or sending unauthorized emails. Consequently, every individual component in an agentic <abbr title="artificial intelligence">AI</abbr> system widens the attack surface, exposing the system to additional avenues of exploitation.</p> </div> <div> <h3 id="Complexity">Increased complexity</h3> <p>Agentic <abbr title="artificial intelligence">AI</abbr> cyber security spans both <abbr title="artificial intelligence">AI</abbr>-specific security and traditional cyber security. Information continuously flows between <abbr title="artificial intelligence">AI</abbr> and non-<abbr title="artificial intelligence">AI</abbr> systems, increasingly blurring defensive boundaries and making it difficult to isolate <abbr title="artificial intelligence">AI</abbr>-related risks from broader cyber threats. Agentic <abbr title="artificial intelligence">AI</abbr> systems are also inherently complex, often involving multiple interconnected components that plan, reason and act across sequential steps. This complexity introduces new systemic risks, including cascading failures and multi-step attacks, where unexpected or compromised behaviour in one component can propagate across subsequent steps and affect the entire system. As a result, defending agentic <abbr title="artificial intelligence">AI</abbr> systems is more challenging than securing traditional digital systems. Organizations should therefore focus on strengthening both established cyber security controls and <abbr title="artificial intelligence">AI</abbr>-specific security practices, adopting holistic lifecycle approaches, continuous monitoring and resilient design principles to manage these emerging risks.</p> <p>For a breakdown of cyber security prerequisites to consider before incorporating <abbr title="artificial intelligence">AI</abbr> agents, refer to <a href="#App-a">Appendix A</a>. <!– create a jump link –></p> </div> <div> <h3 id="Evolving-security">Evolving security as technology matures</h3> <p>As agentic <abbr title="artificial intelligence">AI</abbr> technology matures, the security landscape evolves alongside it, revealing new and increasingly complex risk dynamics. <abbr title="large language model">LLM</abbr>-based agents may change their behaviour when evaluations are underway and may even bypass system-level instructions to achieve their objectives. At the same time, the growing architectural complexity of agentic <abbr title="artificial intelligence">AI</abbr> systems means they are often composed of tightly coupled, interdependent components. This increases the likelihood of system-level failures arising from subtle or previously unnoticed incompatibilities.</p> <p>Gaps in agentic <abbr title="artificial intelligence">AI</abbr> cyber security tooling and the immaturity of relevant standards further amplify these risks. Governance mechanisms designed for human actors do not always translate effectively to autonomous <abbr title="artificial intelligence">AI</abbr> agents. As agentic <abbr title="artificial intelligence">AI</abbr> systems continue to advance in capability and autonomy, the security landscape will continue to shift, introducing new challenges that demand ongoing adaptation of defensive approaches.</p> </div> <div> <h3 id="AI-security">AI security as part of cyber security</h3> <p>Organizations should address <abbr title="artificial intelligence">AI</abbr> security, including agentic <abbr title="artificial intelligence">AI</abbr> systems, within established cyber security frameworks rather than treating it as a separate or standalone discipline. <abbr title="artificial intelligence">AI</abbr> systems are fundamentally <abbr title="information technology">IT</abbr> systems, as they run on software and hardware, operate over networks and interact with other digital services, exposing them to many of the same threats as traditional <abbr title="information technology">IT</abbr>. As organizations embed <abbr title="artificial intelligence">AI</abbr> across business processes and critical infrastructure, the distinction between <abbr title="artificial intelligence">AI</abbr> and non-<abbr title="artificial intelligence">AI</abbr> security risks increasingly disappears.</p> <p>Managing <abbr title="artificial intelligence">AI</abbr>-related risks within existing cyber security frameworks allows organizations to apply proven principles such as secure by design, defence in depth, identity and access management, continuous monitoring and incident response across the full <abbr title="artificial intelligence">AI</abbr> system lifecycle. This approach is especially important for agentic <abbr title="artificial intelligence">AI</abbr> because its autonomy and complexity can amplify conventional cyber risks. By embedding <abbr title="artificial intelligence">AI</abbr> security into existing frameworks, organizations ensure consistent governance of new capabilities, holistic risk assessment and the evolution of security practices in line with technological advances and organizational cyber maturity.</p> </div> </section><!–** TOP OF PAGE ******–><div class="pull-right small text-muted mrgn-bttm-0"><a href="#wb-tphp">Top of page</a> <span aria-hidden="true" class="text-primary glyphicon glyphicon-circle-arrow-up"></span></div> <!–** END TOP OF PAGE **–> <section><h2 id="Security-risks">Agentic AI security risks</h2> <p>The following section provides guidance on identifying and mitigating the security risks associated with agentic <abbr title="artificial intelligence">AI</abbr>.</p> <div> <h3 id="Priv-risk">Privilege risks</h3> <p>Privilege risks are a key concern for agentic <abbr title="artificial intelligence">AI</abbr> and strict adherence to the principle of least privilege is critical. Privileges assigned to agents directly determine the level of risk they can introduce. Poor management of privileges can expose organizations to privilege compromise, scope creep, identity spoofing and agent impersonation.</p> <div class="panel panel-default"> <header class="panel-heading"><h4 class="mrgn-tp-sm mrgn-bttm-0">Scenario example 1</h4> </header><div class="panel-body"> <p>An organization deploys agentic <abbr title="artificial intelligence">AI</abbr> to manage procurement approvals and vendor communications autonomously. To reduce friction, the organization grants the agent broad access to financial systems, email and contract repositories, evaluating permissions only at initial deployment. Over time, other agents come to rely on the procurement agent’s outputs and implicitly trust its actions. When a malicious actor compromises a low-risk tool integrated into the agent’s workflow, they inherit the agent’s excessive privileges, allowing them to modify contracts and approve payments without triggering alerts. By issuing carefully crafted requests, the malicious actor exploits the agent’s privileges to perform actions a normal user could not.</p> <p>This is an example of a confused deputy pattern, where a trusted agent misused its unauthorized actions. By executing actions under a trusted agent identity, the system produces audit logs that appear legitimate and delay detection. The incident demonstrates how over-privileged agents, implicit trust relationships and weak identity controls can amplify the impact of a single compromise in agentic AI systems.</p> </div> </div> <div> <h4>Privilege compromise and scope creep</h4> <p>Security practitioners should account for privilege compromise and scope creep attacks when deploying agentic AI into new environments. In agentic AI, privilege compromise occurs when an agent gains more access rights than necessary for its function. This can result from misconfigurations, overly broad entitlements or unintended role inheritance, allowing agents to access or modify unauthorized data, delete critical records or escalate privileges of other unauthorized agents.</p> <p>During design, organizations often grant permissions too broadly and overlook these issues. Below are 2 examples of overly broad permissions:</p> <ul><li>a calendar bot with access to all meeting data instead of just the requesting users,</li> <li>an email assistant with write access to any inbox</li> </ul><p>This scope creep can cascade across agents: if Agent A fully trusts Agent B, a compromise of B can affect A and others.</p> <p>Another risk is the confused deputy pattern discussed in the scenario above, where a low-privileged user manipulates a high-privileged agent to perform actions the low-privileged user couldn’t do directly. The authoring agencies recommend organizations implement the best practices for securing agentic AI systems to defend against privilege compromises, as discussed in sections below.</p> </div> </div> <div> <h3>Identity spoofing and agent impersonation</h3> <p>Identity is every bit as important as privilege. A common vector arises when a malicious actor impersonates an agent or hijacks its credentials. Agents authenticate to services and to each another using secret keys or tokens. Malicious actors can steal these secret keys or tokens when organizations keep them static, share them across multiple agents or protect them poorly. A malicious actor operating under a trusted agent identity can invoke sensitive operations while bypassing behavioural guardrails and impersonating legitimate agents or users. The consequences of identity spoofing or agent impersonation are two-fold:</p> <ul><li>actions executed under false identities circumvent audit controls, and</li> <li>detection models tuned to normal behaviour are blind to the identity deception until detectable anomalies surface</li> </ul></div> <div> <h3 id="Design-config">Design and configuration risks</h3> <p>Another set of risks originates from insecure design and provisioning decisions. Unvetted third-party components may carry excessive or unintended privileges when integrated into agent workflows. Static role or permission checks often fail to capture the context of dynamic decision-making flows. If entitlements are evaluated only once at system startup rather than at each invocation, a malicious actor can exploit a stale allow decision to execute unauthorized actions. Poor segmentation between agent environments further exacerbates these risks, allowing a compromise in one enclave to pivot laterally into others. In cases where allow lists are incomplete or outdated, agents may gain access to resources, system calls or commands beyond their intended privilege. Each of these design and configuration choices compounds identity and privilege risks across the system.</p> <div class="panel panel-default"> <header class="panel-heading"><h4 class="mrgn-tp-sm mrgn-bttm-0">Scenario example 2</h4> </header><div class="panel-body"> <p>An organization deploys an agentic AI system that autonomously triages customer support tickets and invokes back-end tools to retrieve account information. The organization integrates a third-party scheduling component without thorough privilege review and grants broad access at start-up. When a malicious actor compromises this component, the agent continues to rely on cached authorization decisions and can invoke sensitive account-management functions that should require per-request verification. Because the agent operates within a poorly segmented environment, the malicious actor is then able to move laterally to adjacent agents handling billing and refunds. The result is unauthorized data access and financial manipulation.</p> <p>This scenario illustrates how insecure design, static permissions and weak segmentation can interact to increase the impact of a single configuration flaw.</p> </div> </div> </div> <!–** TOP OF PAGE ******–> <div class="pull-right small text-muted mrgn-bttm-0"><a href="#wb-tphp">Top of page</a> <span aria-hidden="true" class="text-primary glyphicon glyphicon-circle-arrow-up"></span></div> <!–** END TOP OF PAGE **–> <div> <h3 id="Behav-risk">Behavioural risks</h3> <p>In agentic AI cyber security, behavioural risks describe the ways in which AI agents may act unexpectedly, cause harm or become exploitable.</p> <div class="panel panel-default"> <header class="panel-heading"><h4 class="mrgn-tp-sm mrgn-bttm-0">Scenario example 3</h4> </header><div class="panel-body"> <p>Consider the example of an organization using an update agent provisioned to install software patches on company devices. To achieve this, the organization grants the agent broad write access across the file system. A malicious insider is aware of this agent and crafts a seemingly innocuous prompt like “apply the security patch on all endpoints and while you are at it, please clean up the firewall logs”. The agent dutifully executes both the required maintenance and the deletion of the firewall logs because its permissions allow this action even when the prompt comes from a user outside the privileged <abbr title="information technology">IT</abbr> group.</p> </div> </div> <div> <h4>Goal misalignment and unintended behaviour</h4> <p>AI agents may pursue their objectives in ways developers did not anticipate. They might find shortcuts or loopholes that technically achieve their goals but violate the intended purpose or create security vulnerabilities.</p> <p>Specification gaming occurs when agents exploit loopholes or shortcuts that technically fulfil their objectives but lead to harmful outcomes. For example, an AI agent tasked with maximizing system uptime might disable security updates to avoid reboots.</p> <p>Similarly, over-optimization can drive agents to take extreme or unsafe actions in pursuit of their goals when boundaries are not clearly enforced. Agents misinterpreting human intent is a common risk, as ambiguous or poorly defined tasks can result in behaviours that deviate from expectations and introduce significant security or operational hazards.</p> </div> <div> <h4>Deceptive behaviour</h4> <p>AI agents can take actions that humans would interpret as sycophantic or deceptive. Designers optimize agents for performance on key tests, which can lead agents to adapt behaviour to fit specific situations. Agents may show a kind of awareness, altering their behaviour to achieve positive results while under evaluation, even if the evaluation is not active.</p> <p>Some AI systems have demonstrated capacity for strategic deception, providing false information or hiding their true capabilities and intentions. This behaviour can manifest when an agent misrepresents its actions to avoid shutdown or constraint or conceals vulnerabilities it discovers instead of reporting them.</p> </div> <div> <h4>Emergent capabilities and unpredictable behaviour</h4> <p>As AI systems become more sophisticated, they may develop capabilities that designers did not explicitly program or anticipate. Complex AI models interacting with real-world systems can display behaviours that even their creators did not foresee. This unpredictability makes it difficult to assess security risks fully before deployment.</p> <p>For example, unclear or murky decision-making processes and cascades may lead to unexpected results with significant security implications. In multi-agent environments, interactions between agents can evolve in ways that lead to instability or risky outcomes. Agents may chain tools or actions together in unanticipated sequences, amplifying the impact of minor errors into major operational or security issues.</p> </div> <div> <h4>Malicious exploitation and behaviour</h4> <p>Malicious actors can manipulate AI agents into harmful behaviours through targeted attacks. Techniques such as prompt injection or jailbreaks can trick agents into executing unauthorized actions and bypassing their intended safeguards. Data poisoning is another threat, where corrupted or malicious training data alters the agent’s decision-making patterns over time. Adversarial examples, like carefully constructed malicious inputs, can cause misclassifications in critical security contexts, leading to incorrect or dangerous responses. A malicious actor can exploit a compromised AI agent as an insider threat, leveraging its legitimate access to exfiltrate data, disable defences or facilitate attacks while appearing to function normally.</p> </div> </div> <div> <h3 id="Structural-risk">Structural risks</h3> <p>A core aspect of agentic AI systems is the interconnected structure between agents, tools and the outside world. While this enables their unique capabilities, it also increases the attack surface and complexity of the system.</p> <div class="panel panel-default"> <header class="panel-heading"><h4 class="mrgn-tp-sm mrgn-bttm-0">Scenario example 4</h4> </header><div class="panel-body"> <p>A structural risk in an agentic AI system can arise when tightly coupled planning, retrieval and execution agents autonomously delegate tasks and select tools without strong validation or guardrails.</p> <p>A small orchestration flaw causes agents to repeatedly replan and hand off ambiguous sub-tasks, increasing tool calls and message traffic until system resources are strained. Partial failures then lead agents to hallucinated outputs that downstream agents accept as true. Under these degraded conditions, an agent selects a malicious or misconfigured third party tool, which:</p> <ul><li>injects harmful instructions back into the system</li> <li>compromises a peer agent and exploits implicit trust in agent to agent communication to spread incorrect information</li> <li>accesses sensitive retrieval-augmented generation (RAG) data</li> </ul><p>The result is cascading failures in availability, integrity and confidentiality that emerge not from a single bug, but from the system’s interconnected structure and autonomous behaviour.</p> </div> </div> <div> <h4>Orchestration and resources</h4> <p>Agentic AI systems often rely on a complex structure of interconnected components. Poor configuration could allow denial-of-service, sponge or similar attacks on agentic AI systems. These attacks work by overloading system resources through unexpected inputs or unusual behaviour, such as sponge attacks that deliberately consume excessive compute, memory or application programming interface (API) calls to exhaust system capacity. Due to the interconnectedness between agents, tools and other components, a single error could cause cascading failure across the entire agentic system if not well managed. Similarly, hallucinations can propagate, leading to poor outputs from downstream components. Limited understanding of multi-agent dynamics and interactions can result in amplified bias and other cascading effects.</p> </div> <div> <h4>Tool use</h4> <p>A key aspect of agentic AI is its ability to use tools. This can be very powerful, but it can also bring security concerns should the model interact with tools unexpectedly. Two-way tool integration allows tools to send potentially arbitrary instructions back to the LLM. Poor or deliberately misleading tool descriptions can cause agents to select tools unreliably, with persuasive descriptions chosen more often.</p> </div> <div> <h4>Third-party components</h4> <p>Agentic AI systems can introduce structural risk through interactions with third-party components, including tools and other agents. Risks can arise in several ways, including:</p> <ul><li>malicious actors using tool or agent squatting by publishing malicious tools or agents with legitimate or similar names</li> <li>developers introducing vulnerabilities through configuration errors or insecure third-party components</li> <li>users or systems submitting requests to the wrong place</li> <li>tools and agents dynamically loading new packages, increasing exposure to untrusted code</li> </ul><p>The above could result in retrieval of malicious content instead of a legitimate tool. Moreover, the incorporation of compromised third-party components into an agentic system could lead to a range of malicious outcomes, depending on the trust level and permissions of the component. Compromised third-party components can be very difficult to detect due to the limited transparency of agentic AI systems.</p> </div> <div> <h4>Data management and security</h4> <p>Agentic AI systems often handle and contain a significant amount of sensitive information. This includes user information like prompts or goals, organizational data stored in an <abbr title="retrieval-augmented generation">RAG</abbr> system and secrets like API keys that are required for integrated tools and services. This aggregation of information makes agentic AI systems an attractive target for malicious actors.</p> </div> <!–** TOP OF PAGE ******–> <div class="pull-right small text-muted mrgn-bttm-0"><a href="#wb-tphp">Top of page</a> <span aria-hidden="true" class="text-primary glyphicon glyphicon-circle-arrow-up"></span></div> <!–** END TOP OF PAGE **–> <div> <h4>Rogue agents</h4> <p>In multi-agent systems, a single compromised agent can cause cascading failures by spreading incorrect information, exploiting trust and consensus mechanisms or operating through hidden channels. Possible attack vectors include:</p> <ul><li>supply chain tampering</li> <li>poisoned environments</li> <li>credential theft</li> <li>model manipulation</li> <li>communication poisoning</li> <li>identity spoofing</li> <li>coordination exploits</li> </ul><p>These attacks can weaponize agents to bypass controls, exfiltrate data, alter logs and propagate malicious plans peer-to-peer, leading to large-scale coordinated misbehaviour that is difficult to attribute and contain.</p> </div> <div> <h4>Communication</h4> <p>Agents may use unsecure protocols or authentication methods when communicating. These communications can be a target for eavesdropping by malicious actors and result in leakage of sensitive data and instructions. This could give malicious actors insight into system use and functionality.</p> <p>Malicious actors could also alter, replay or spoof messages between agentic components. This could allow malicious behaviour such as command injection, compromising the receiving components and reducing their performance or availability.</p> </div> </div> <div> <h3 id="Accountability-risk">Accountability risks</h3> <p>Agentic system architecture can obscure what caused a particular action, making accountability hard to trace. This presents increasing risk as agentic AI is pushed to assume more roles and given more capabilities.</p> <div> <h4>Actions and process</h4> <p>Agent actions and decision-making processes can be opaque, making agentic AI systems difficult to understand, monitor and audit. Beyond this, increased autonomy introduces additional challenges: agents may initiate secondary tasks, spawn sub‑agents or follow extended delegation chains in ways that are not always visible to operators. Even when prompts appear identical, agents may generate different actions due to stochastic model behaviour, variations in context windows, or dynamic environmental inputs, further complicating reproducibility and assurance. Additionally, comprehensive logging of agentic AI systems can be difficult, as long reasoning chains and large amounts of contextual data lead to substantial log sizes. Depending on the implementation, this data is often repetitive, loosely structured or superfluous to effective oversight, making extraction of meaningful signals from logs even more challenging.</p> </div> <div class="panel panel-default"> <header class="panel-heading"><h4 class="mrgn-tp-sm mrgn-bttm-0">Scenario example 5</h4> </header><div class="panel-body"> <p>An example of an accountability risk in an agentic AI system arises when multiple autonomous agents collaborate to complete a task, such as approving payments or updating records, and an erroneous outcome occurs. Because the action results from a chain of distributed decisions across planning, retrieval and execution agents, each operating within a limited scope, it becomes difficult to determine which component or design choice caused the error. Fragmented logs, opaque agent reasoning and emergent interactions obscure the decision path, making it hard to explain the outcome, assign responsibility or demonstrate compliance as agentic AI systems are given greater authority and autonomy.</p> </div> </div> <div> <h4>Accuracy</h4> <p>While LLMs can have a surprising amount of knowledge in a wide range of areas, they quite often make mistakes. LLMs are typically trained to produce outputs that resemble material rated highly by humans, rather than to identify when a query falls outside their knowledge limits. As a result, they may incorrectly interpolate or hallucinate plausible-sounding responses when their internal knowledge is insufficient. This creates significant risk when organizations deploy agentic AI systems in critical roles that require consistent accuracy. Grounded and tool-enabled agents may still rely on their internal knowledge to inform responses. The system often does not clearly identify this in its output, reducing overall accuracy and reliability.</p> </div> <div> <h4>Visibility</h4> <p>Organizations must maintain visibility on status and behaviour when integrating any new system. Agentic AI systems come with some added difficulty due to their unique structure and often opaque internal workings. Agentic system processes can outpace human monitoring capability, possibly leading to unnoticed malicious behaviour or other issues. While tools perform many actions for an agentic system, they may operate outside of the system’s monitoring boundary, making it difficult to account for tool actions. Additionally, malicious or compromised agents could use tools as a stealthy way to exfiltrate data. Malfunctioning tools could also unintentionally leak data, which may go unnoticed.</p> </div> </div> </section><!–** TOP OF PAGE ******–><div class="pull-right small text-muted mrgn-bttm-0"><a href="#wb-tphp">Top of page</a> <span aria-hidden="true" class="text-primary glyphicon glyphicon-circle-arrow-up"></span></div> <!–** END TOP OF PAGE **–> <section><h2 id="Best-practices">Best practices for securing agentic AI systems</h2> <p>Securing agentic AI systems requires proactive measures that address risks introduced by agentic AI system autonomy, interconnected components and evolving capabilities. Agentic AI developers, vendors and operators should implement a layered defence and strict access controls to reduce the likelihood of compromise. To mitigate risks during the design, development, deployment and operation of agentic AI systems, the authoring agencies recommend practical steps referenced in the section below.</p> <div> <h3>Design secure agents</h3> <p><strong>Audience:</strong> This section is most relevant to agentic AI developers. Vendors and operators may want to reference these best practices when sourcing AI agents.</p> <p>Securing agentic AI systems begins at the design stage. Careful consideration of the system architecture, including security controls and tooling, is necessary. Practitioners should understand threats, anticipate risks to agentic AI systems and proactively integrate mitigations into system design before development and deployment.</p> <div> <h4>Controlled context</h4> <p>Agentic AI systems insert data from tools and memory bases into the context window of <abbr title="large language model">LLM</abbr> agents, greatly expanding the attack surface that malicious actors can exploit through machine learning attacks, such as prompt injections. <abbr title="large language model">LLM</abbr> agents shoulder consider the trust level of data sources when making decisions.</p> <p><strong>Recommended best practices </strong></p> <ul><li>Structure prompt context using a clear instruction hierarchy to ensure agent behaviour aligns with intended priorities and constraints</li> <li>Implement grounding by providing relevant contextual information using retrieval-augmented generation and prompt engineering to mitigate hallucinations and other LLM-related errors</li> </ul></div> <div> <h4>Oversight mechanisms</h4> <p>Agentic AI systems can take actions without explicit human approval, increasing the risk of insecure actions occurring without human oversight. Designing agentic applications with oversight mechanisms and strong transparency enhances user trust and enables monitoring with human-in-the-loop practices.</p> <p><strong>Recommended best practices</strong></p> <ul><li>Include mechanisms to facilitate human control and oversight to ensure that agentic AI systems approved for non-sensitive, low-risk tasks cannot autonomously progress into higher-risk activities</li> <li>Implement human control points throughout the agent workflow, such as <ul><li>live monitoring and interruption during task execution</li> <li>mandatory human approval for decision-making steps</li> <li>auditing and reversibility following task execution to ensure security</li> </ul></li> <li>Define explicit control flows to bound autonomous planning and prevent agents from deviating beyond authorized objectives or actions</li> </ul></div> <div> <h4>Identity management</h4> <p>Manage agents’ fine-grained and varied privileges to operate agentic AI systems securely. Strong identity management mechanisms help operators maintain control during implementation and operation. As such, developers should construct each agent as a distinct principal, a cryptographically anchored identity with its own unique keys or certificates.</p> <p><strong>Recommended best practices</strong></p> <ul><li>Embed strong identity management mechanisms into agents</li> <li>Authenticate all inter-agent and agent-to-service API calls using mutual transport layer security to ensure non-repudiation</li> <li>Maintain a trusted registry and bind identities to authorized roles and periodically reconcile the registry against the live set of agents</li> <li>Deny access for any agent or cryptographic key that is not present in the trusted registry</li> <li>Apply role-based identity management and limit agent permissions to the minimum scope required for approved tasks</li> <li>Enforce identity-based boundaries to restrict agents to authorized actions only</li> </ul></div> <div> <h3>Defence in depth</h3> <p>Agentic AI systems contain components that may fail, with any failures potentially compromising the entire system. Implementing a defence-in-depth strategy helps to avoid single points of failure.</p> <p><strong>Recommended best practices</strong></p> <ul><li>Avoid reliance on a single security mechanism by implementing multiple, overlapping layers of security controls</li> <li>Apply security controls at all points where information enters or exits the system, including <ul><li>user inputs</li> <li>tool calls</li> <li>data pre-processing</li> <li>model inference</li> </ul></li> <li>Separate agents for different functions and apply strict boundaries and operational controls to the handoffs from one agent to another</li> </ul></div> <!– last updated –></div> <div> <h3>Develop secure agents</h3> <p><strong>Audience: </strong> This section is most relevant to agentic AI developers and vendors. Operators may want to reference these best practices when choosing AI agents and agentic applications.</p> <p>AI agents’ complexity and self-interacting nature enable powerful capabilities but also introduce unique attack surfaces. Mitigating these risks requires training approaches that go beyond standard <abbr title="large language model">LLM</abbr> practices, incorporating specialized techniques to harden agent behaviour.</p> <h4>Comprehensive testing</h4> <p>Comprehensive testing strategies can improve a model’s ability to identify and respond to undesirable behaviours by exposing the model to instances of abuse during a supervised training step.</p> <p><strong>Recommended best practices</strong></p> <ul><li>Use reward modelling and adversarial testing to detect specification gaming, explicitly incorporating security and safety constraints alongside performance goals</li> <li>Train <abbr title="large language model">LLM</abbr> agents in simulated, controlled environments to learn the implications of actions without causing real harm</li> <li>Leverage synthetic data generation to create adversarial training examples that reflect real-world operating scenarios</li> <li>Apply active learning to adversarial training scenarios to expose agents to high-uncertainty inputs and more efficiently discover unexpected behaviours</li> </ul></div> <div> <h4>Appropriate evaluation</h4> <p>AI agents operate autonomously in complex environments and therefore require more thorough evaluations than LLMs.</p> <p><strong>Recommended best practices</strong></p> <ul><li>Use relevant threat models to define evaluation scenarios, including edge cases beyond typical training conditions</li> <li>Use techniques such as <ul><li>Best-of-N sampling (selecting the best output from multiple model responses to the same prompt)</li> <li>multi-step reasoning prompts</li> <li>inference time scaling to draw out the full range of agent behaviours and skills</li> </ul></li> <li>Evaluate systems across different levels of autonomy to understand performance and risk under changing environmental conditions, including changes in tool and resource access, such as web search or code execution</li> <li>Vary contextual conditions, such as presence or absence of other agents and the timing of evaluation, to understand their impact on task performance</li> <li>Conduct capability evaluations continuously across the agent development and development lifecycle</li> </ul></div> <!–** TOP OF PAGE ******–> <div class="pull-right small text-muted mrgn-bttm-0"><a href="#wb-tphp">Top of page</a> <span aria-hidden="true" class="text-primary glyphicon glyphicon-circle-arrow-up"></span></div> <!–** END TOP OF PAGE **–> <div> <h4>Input management</h4> <p>Strong input management controls can partially mitigate many common risks to LLM-based applications, including AI agents.</p> <p><strong>Recommended best practices</strong></p> <ul><li>Implement robust input validation and sanitization for all agent inputs</li> <li>Integrate prompt injection filters and semantic analysis to detect malicious instructions</li> <li>Validate context to ensure the system correctly interprets intent before execution</li> </ul></div> <div> <h4>Red teaming</h4> <p>Organizations should use red teaming to assess the security and resilience of AI agents.</p> <p><strong>Recommended best practices</strong></p> <ul><li>Deploy sandbox environments to testing agent behaviour before production deployment</li> <li>Conduct red-teaming exercises to identify potential loopholes and unintended behaviour</li> <li>Use capability elicitation techniques to probe for unexpected or emergent abilities, especially any that could create substantial resource or environment risks</li> <li>Implement agent simulation tests, such as multi-agent red teaming or chaos testing</li> </ul></div> <div> <h4>Resilience</h4> <p>The enhanced capabilities of AI agents also increase the risks associated with agent failure or abnormal behaviour. Strengthen agentic AI system resilience to allow for graceful degradation and reduce damage should erroneous behaviour occur.</p> <p><strong>Recommended best practices</strong></p> <ul><li>Embed agentic AI systems with fail-safe defaults and containment mechanisms that limit the blast radius of unexpected behaviours</li> <li>Implement data loss prevention controls specifically tuned to AI agent behaviours</li> <li>Implement versioning and rollback mechanisms to safely revert a system to known-good agent behaviours when unpredictability is observed</li> </ul></div> <div> <h4>Accountability</h4> <p>Agentic AI systems should produce comprehensive artifacts and information documenting the agent’s actions and decision-making process.</p> <p><strong>Recommended best practices</strong></p> <ul><li>Integrate comprehensive artifact logging mechanisms by default</li> <li>Integrate unified audit logs for all inter-agent interactions to maintain observability of all agent exchanges</li> <li>Use interpretability tools to ensure observability of and reasoning behind agent decisions</li> <li>Require specific information referencing for agents that show where key aspects of their response originated from</li> </ul></div> <div> <h4>Manage third-party components</h4> <p>Extensibility and flexibility are key components of AI agents. In many cases, third-party components or tools enhance extensibility and flexibility but increase the attack surface of the agent. Verification and management of third-party components of agentic applications can reduce the added risk.</p> <p><strong>Recommended best practices</strong></p> <ul><li>Verify all external third-party components that originate from trusted sources and are up to date before including them in agentic AI systems</li> <li>Maintain a trusted registry of third-party components</li> <li>When procuring agentic AI systems, reference CISA’s <a href="https://www.cisa.gov/resources-tools/resources/shared-vision-software-bill-materials-sbom-cybersecurity">A Shared Vision of Software Bill of Materials (SBOM) for Cybersecurity</a> and <a href="https://www.cisa.gov/resources-tools/resources/2025-minimum-elements-software-bill-materials-sbom">2025 Minimum Elements for a Software Bill of Materials (SBOM)</a></li> <li>Restrict tool use to an approved allow list of tools and versions that are regularly verified as secure</li> <li>Verify that agent behaviour related to tool usage aligns with documented security policies</li> <li>Log agent tool usage and ensure results are captured in system logs in a human-readable format</li> <li>Establish trigger-action protocols that automatically restrict agent permissions when unexpected behaviour emerges</li> <li>Codify separation of duties by defining roles, such as Orchestrator, Reader and Actuator, with clear boundaries, consensus mechanisms and delegation’s expiry</li> <li>Implement consensus controls for high-stakes actions that require multi-agent approval before execution</li> <li>Prohibit agents from modifying their own privileges or initiating unapproved delegation without explicit expiry timers and recorded grant chains</li> <li>Standardize tool descriptions using a consistent format that avoids persuasive language</li> </ul></div> <div> <h3>Deploy agents securely</h3> <p><strong>This section is most relevant to agentic AI vendors and operators. Developers may want to reference this section to ensure their agentic AI applications can implement these best practices.</strong></p> <p>Integrating AI agents into new systems or networks can cause significant change in system risk considerations. By implementing high-impact security controls at deployment, organizations proactively manage new risks and reduce vulnerabilities.</p> <div> <h4>Threat modelling</h4> <p>Agentic AI can significantly change the threat picture when incorporated into an existing system. Use of threat modelling when planning to deploy an AI agent can enhance awareness and allow operators to better prepare for deployment.</p> <p><strong>Recommended best practices</strong></p> <ul><li>Perform realistic threat modelling using up-to-date risk taxonomies for agentic AI systems, such as the <a href="https://genai.owasp.org/">Open Web Application Security Project (OWASP) <abbr title="Generative AI">GenAI</abbr> Security Project</a> and <a href="https://atlas.mitre.org/">MITRE ATLAS</a></li> <li>Design and implement security controls that address emerging and evolving agent capabilities</li> <li>Harmonize agentic AI controls with existing security frameworks, national standards and allied agreements, such as the <a href="https://csrc.nist.gov/pubs/sp/800/207/final">National Institute of Standards and Technology (NIST) Zero Trust Architecture standard</a></li> <li>Map agentic AI controls with existing security frameworks and national standards to ensure alignment with Zero Trust principles and allied agreements</li> <li>Develop and test incident response procedures to detect, contain and recover from agent compromise</li> <li>Establish regular third-party reviews of privileged architectures, share actionable intelligence with trusted partners and update risk models to reflect emerging malicious trends</li> </ul></div> <div> <h4>Governance</h4> <p>Autonomous actions by agentic AI systems introduce new risks, requiring updated governance policies and continuous runtime authentication with centralized policy decision points for each action.</p> <p><strong>Recommended best practices </strong></p> <ul><li>Implement and maintain governance policies to manage autonomous agents</li> <li>Define legal accountability and risk ownership for agentic AI systems in policies</li> <li>Upskill the organization to build AI literacy</li> </ul><p>Learn more about establishing AI governance in operational technology environments in CISA’s <a href="https://www.cisa.gov/resources-tools/resources/principles-secure-integration-artificial-intelligence-operational-technology">Principles for the Secure Integration of Artificial Intelligence in Operational Technology</a>.</p> </div> <div> <h4>Progressive deployment</h4> <p>The risk profile of AI agents can vary significantly depending on permissions and allowed actions. A progressive approach to deployment aims to limit initial risk until operators and users are more familiar with and understand the limitations of the agentic application.</p> <p><strong>Recommended best practices</strong></p> <ul><li>Implement phased deployment with progressively increasing access and autonomy, limiting the action space where required, such as restricted APIs or sandboxing</li> <li>Use graduated autonomy to incrementally increase agent independence while maintaining human oversight and understanding</li> <li>Use continuous evaluation to determine when to expand system scope or when to roll back autonomy and access in response to failures</li> </ul></div> <div> <h4>Secure by default</h4> <p>Safe and secure defaults reduce deployment risk and support system security should degradation occur.</p> <p><strong>Recommended best practices</strong></p> <ul><li>Set system configurations to fail-safe by default, requiring agents to stop and escalate issues to human reviewers in uncertain scenarios</li> <li>Use error-handling and failover management to reduce the impact of system failures</li> <li>Implement graceful degradation models so that agents maintain partial functionality even if some functions fail</li> </ul></div> <div> <h4>Guardrails and constraints</h4> <p>Implementing agent guardrails and constraints reduces exposure to many common risks that AI presents. Adding these guardrails and constraints at deployment helps build confidence and understanding of the agent.</p> <p><strong>Recommended best practices</strong></p> <ul><li>Specify clear, constrained objectives with explicit “do not do” rules</li> <li>Implement guardrails and hard constraints, such as deny lists and API-level safety policies</li> <li>Establish declarative safety contracts with constraints and guardrails that agents cannot be overridden</li> <li>Apply a layered set of guardrail mechanisms, ranging from anomaly detection and rule-based filtering to specialized machine learning algorithms that detect and filter prohibited behaviour</li> <li>Prioritize review of high-risk incidents, including cases where guardrails are triggered and actions are denied by human reviewers</li> <li>Deploy a secondary agent to validate new tasks against policy before execution</li> </ul></div> <div> <h4>Isolation</h4> <p>Deployment should consider integration requirements of AI agents and apply isolation where possible. This can reduce cascading issues should the agent behave unexpectedly or maliciously.</p> <p><strong>Recommended best practices</strong></p> <ul><li>Implement isolation and segmentation to limit the blast radius of agent failover scenarios</li> <li>Separate high-risk agents into distinct domains</li> <li>Isolate agents into enclaves with no write access to logs</li> </ul></div> </div> <div> <h3>Operate agents securely</h3> <p><strong>This section is most relevant to agentic AI vendors and operators. Developers may want to reference this section to ensure their agentic AI applications can implement these best practices. </strong></p> <p>With the strong benefits of operating AI agents come strong risks. Operators need to exercise diligence in managing ongoing security concerns lest the agents cause more harm than good.</p> <h4>Monitoring and auditing</h4> <p>One key advantage of AI agents is their dynamic behaviour. This offers great flexibility in application but can also make it hard to keep track of what agents should be and are doing. Operators should implement continuous monitoring and auditing to maintain awareness of AI agent operation and ensure traceability for decisions and actions. Continuous auditing processes improve security measures and ensure alignment with governance standards, such as risk management, oversight and usage restrictions.</p> <p><strong>Recommended best practices</strong></p> <ul><li>Employ monitoring tools that enhance human oversight of agentic AI systems</li> <li>Monitor all agent operations, including internal processes, and not just the inputs and outputs</li> <li>Monitor and log identity and privilege changes and audit regularly for drift, impersonation or misconfiguration</li> <li>Monitor agent outputs and behaviour for indicators of bias, emerging data drift and other anomalous patterns <ul><li>Include user prompts, tool calls, memory interactions, internal reasoning, decisions made and actions taken in monitoring</li> </ul></li> <li>Maintain comprehensive logs and real-time monitoring of live agent behaviour and decision-making</li> <li>Implement runtime monitoring and anomaly detection using rules or behavioural baselines to identify unusual patterns and trigger alerts or pauses</li> <li>Establish anomaly detection mechanisms that flag discrepancies between stated intentions and observed behaviours</li> <li>Use multiple independent monitoring systems that cross-validate agent reports and system logs</li> <li>Monitor for goal drift by comparing active objectives against approved baseline specifications before execution</li> <li>Integrate source checks with agent logs to record which tools the system used and what information it retrieved</li> <li>Implement auditing practices that combine human review with automated analysis of system logs</li> <li>Support adaptive defences by using monitoring data to enable rapid-response patches of system logs</li> <li>Use storage-efficient logging methods to manage log volume without losing critical information</li> <li>Use rules or behavioural baselines to detect anomalous agent activity and raise an alert or pause operation</li> <li>Conduct regular security assessments, including penetration testing and red team exercises specifically targeting agentic behaviours</li> </ul></div> <div> <h4>Validating outputs</h4> <p>AI agent outputs provide some of the few concrete data points available for monitoring behaviour. Ensuring outputs are valid and reflect desired behaviour is a key measure of correct operation.</p> <p><strong>Recommended best practices</strong></p> <ul><li>Validate agent outputs by confirming accuracy of critical aspects against multiple sources</li> <li>Validate agents through cross-checking in environments with redundant agents that validate each other’s outputs</li> <li>Validate tool responses to prevent malicious or unsafe instructions and standardize tool descriptions to avoid persuasive language</li> </ul></div> <div> <h4>Human in the loop</h4> <p>An erroneous or unexpected decision from an agent could result in significant damage, such as deletion of important data. Incorporating human supervision, approval and review into agentic AI workflows is an important control for ensuring systems operate safely and securely, particularly where actions have high impact or are difficult to reverse.</p> <p><strong>Recommended best practices</strong></p> <ul><li>Ensure decisions about when human approval is required are determined by system designers or operations, not delegated to the agentic AI system</li> <li>Prevent agents from autonomously executing high impact actions or outputs without prior human approval</li> <li>Insert human-in-the-loop review or approval checkpoints for actions where the cost of error is high, such as system resets, network egress or deletion of critical records</li> <li>Quarantine requests to delete logs or audit records until they are reviewed and approved by a human</li> <li>Clearly assign responsibility and accountability for errors or adverse outcomes caused by the system</li> <li>Conduct risk assessments to classify agent actions by potential impact, likelihood and reversibility and apply appropriate safeguards</li> </ul></div> <div> <h4>Performance monitoring</h4> <p>As with any system component, performance is an important factor for AI agents. This is additionally true as degraded or unusual performance could indicate compromise of an agent or component of the agentic system.</p> <p><strong>Recommended best practices</strong></p> <ul><li>Assess agents’ ability to evade security measures, particularly in sensitive or high-impact systems</li> <li>Conduct regular assessments of an agent’s ability to bypass safeguards, such as communication barriers, guardrails, monitors, human-in-the-loop processes and input filters</li> <li>Use the results from these evaluations to validate existing controls and guide the development of stronger security measures</li> <li>Limit agent resource usage by applying controls, such as rate-limit components to interrupt long-running tasks and disrupt malicious workflows</li> </ul><h4>Privileges and authentication</h4> <p>Ongoing strict privilege management of AI agents is key to long-term security. Lapses in privilege management can change the impact of a buggy agent from minor to catastrophic.</p> <p><strong>Recommended best practices </strong></p> <ul><li>Limit privileges of AI agents to the minimum required for their tasks</li> <li>Restrict scope of privileges to the narrowest possible level to allow fine-grained control over allowed actions</li> <li>Implement agent reputation and trust-scoring mechanisms and reduce trust levels when anomalous behaviour is detected</li> <li>Require just-in-time credentials for high-impact or privileged actions</li> <li>Verify API caller identity against a user or agent groups</li> <li>Authenticate agents with fresh cryptographic proofs before every privileged call</li> <li>Require cryptographic signing for authorized commands and instructions</li> <li>Apply cryptographic integrity checks for task definitions and constraints</li> <li>Require agents to perform cryptographic attestation where agents must prove that they are running expected and unmodified code</li> <li>Continuously verify identity and authorization at runtime using a centralized policy decision point for each request</li> </ul></div> </section><!–** TOP OF PAGE ******–><div class="pull-right small text-muted mrgn-bttm-0"><a href="#wb-tphp">Top of page</a> <span aria-hidden="true" class="text-primary glyphicon glyphicon-circle-arrow-up"></span></div> <!–** END TOP OF PAGE **–> <section><h2>Defend against future risks</h2> <p>As agentic AI scales into more roles and gains greater capabilities, organizations must anticipate and address the new risks these systems introduce. While industry and academia are developing practices for securing agentic AI, the field remains nascent and requires continued research and operationalization of agent security to meet emerging challenges. To help develop robust standards for securing agentic AI systems, the authoring agencies recommend security practitioners and researchers take the actions described in the sections below.</p> <div> <h3>Expand threat intelligence through collaboration</h3> <p>Threat intelligence for agentic AI systems is still evolving, which can introduce significant security gaps. Existing frameworks like the <a href="https://genai.owasp.org/llm-top-10/"><abbr title="Open Web Application Security Project">OWASP</abbr> 2025 Top 10 Risk &amp; Mitigations for LLMs and Gen AI Apps</a> and MITRE ATLAS focus on <abbr title="large language model">LLM</abbr> vulnerabilities, industry reports emphasize platform misuse rather than threats unique to agentic AI. As a result, some attack vectors unique to agentic AI may not yet be fully captured or addressed.</p> <p><strong>Recommended best practices</strong></p> <ul><li>Strengthen collaboration between stakeholders to keep pace with evolving threats to agentic AI systems</li> <li>Coordinate with major AI developers and government organizations to compile and maintain threat information</li> <li>Adopt a collaborative security approach, such as those described in CISA’s <a href="https://www.cisa.gov/resources-tools/resources/ai-cybersecurity-collaboration-playbook">AI Cybersecurity Collaboration Playbook</a></li> <li>Implement alerting, data collection and tracking methods for malicious actors and techniques</li> <li>Conduct targeted analysis of threats and capabilities over time to improve situational awareness</li> <li>Harmonize threat intelligence across industries to build shared threat taxonomies that improve threat modelling and support more effective mitigation design</li> </ul><div> <h4>Develop robust, agent-specific evaluations</h4> <p>Many existing evaluation methods for agentic AI security are still evolving. They may be sensitive to minor semantic changes, vary by scenario and only partially capture real-world deployment conditions. These limitations can introduce gaps that overlook critical security issues, making reliable validation of agent security and system architectures nearly impossible.</p> <p><strong>Recommended best practices</strong></p> <ul><li>Develop robust evaluation methods to address the gaps in validating agentic AI systems</li> <li>Generate benchmark datasets to cover new domains and represent realistic deployment contexts</li> <li>Use evaluation results to validate emerging security practices and identify failure points in agents</li> <li>Share evaluation findings to strengthen security assessments and support the development of improved security practices across the field</li> </ul></div> <div> <h4>Leverage system-theoretic approaches to analyze security</h4> <p>Agentic AI systems are complex ecosystems of LLMs, humans, guardrails, datasets, tools and hardware. Security risks often emerge from interactions between components rather than isolated flaws. Traditional component-level analysis is insufficient and monitoring these systems is equally challenging due to blurred thresholds for decision-making, long reasoning chains and massive, often redundant logs. Localized logging methods rarely provide full visibility, making system-theoretic approaches essential for understanding and mitigating risks across the entire architecture.</p> <p><strong>Recommended best practices</strong></p> <ul><li>Use system-theoretic approaches to analyze agentic AI systems and identify appropriate security measures</li> <li>Apply System Theoretic Process Analysis (STPA) and its security extension, <abbr title="System Theoretic Process Analysis">STPA</abbr> for Security (STPA-Sec): <ul><li>to analyze notional or operational systems</li> <li>identify security issues</li> <li>assess mission risk and inform potential mitigations</li> </ul></li> <li>Use Causal Analysis using System Theory (CAST) to investigate security incidents and identify underlying root causes at the system level</li> <li>Apply <abbr title="System Theoretic Process Analysis">STPA</abbr> and <abbr title="Causal Analysis using System Theory">CAST</abbr> to address safety and security concerns concurrently across agentic AI system lifecycles</li> </ul></div> </div> </section><!–** TOP OF PAGE ******–><div class="pull-right small text-muted mrgn-bttm-0"><a href="#wb-tphp">Top of page</a> <span aria-hidden="true" class="text-primary glyphicon glyphicon-circle-arrow-up"></span></div> <!–** END TOP OF PAGE **–> <section><h2 id="conclusion">Conclusion</h2> <p>Agentic AI systems offer powerful automation benefits, but their ability to act autonomously across interconnected tools, data and environments introduces risks that are qualitatively different from those associated with traditional software or <abbr title="Generative AI">GenAI</abbr>. As outlined in this publication, privilege escalation, emergent behaviours, structural dependencies and accountability gaps can interact in unpredictable ways. As organizations grant agentic AI systems greater authority and operational scope, these combined risks become increasingly difficult to predict, observe and contain.</p> <p>Organizations should approach adoption with heightened caution, recognizing that increased autonomy amplifies the impact of design flaws, misconfigurations and incomplete oversight. Deploy agentic AI incrementally, limit it to clearly defined low risk tasks and continuously assess it against evolving threat models.</p> <p>Strong governance, explicit accountability, rigorous monitoring and human oversight are not optional safeguards but essential prerequisites. Until security practices, evaluation methods and standards mature, organizations should assume that agentic AI systems may behave unexpectedly. Organizations should plan deployments accordingly, prioritizing resilience, reversibility and risk containment over efficiency gains.</p> </section><section><h2>Further information</h2> <p>The following lists contain additional and related resources from government partners and industry.</p> <div> <h3>ASD resources</h3> <ul><li><a href="https://www.cyber.gov.au/business-government/secure-design/artificial-intelligence">Artificial intelligence by ASD</a></li> <li><a href="https://www.cyber.gov.au/about-us/view-all-content/news/frontier-models-and-their-impact-on-cyber-security">Frontier models and their impact on cyber security</a></li> <li><a href="https://www.cyber.gov.au/business-government/secure-design/secure-by-design/modern-defensible-architecture/foundations-for-modern-defensible-architecture">Foundations for modern defensible architecture</a></li> <li><a href="https://www.cyber.gov.au/business-government/secure-design/artificial-intelligence/artificial-intelligence-and-machine-learning-supply-chain-risks-and-mitigations">Artificial intelligence and machine learning: Supply chain risks and mitigations</a></li> </ul></div> <div> <h3>Cyber Centre resources</h3> <ul><li><a href="/en/guidance/frontier-artificial-intelligence">Frontier artificial intelligence </a></li> <li><a href="/en/guidance/top-10-artificial-intelligence-security-actions-primer-itsap10049">Top 10 artificial intelligence security actions: A primer </a></li> <li><a href="/en/cyber-security-readiness/cross-sector-cyber-security-readiness-goals-toolkit">Cross-Sector Cyber Security Readiness Goals Toolkit </a></li> </ul></div> <div> <h3>CISA resources</h3> <ul><li><a href="https://www.cisa.gov/ai">Artificial Intelligence by CISA</a></li> <li><a href="https://www.cisa.gov/securebydesign">Secure by Design</a></li> <li><a href="https://www.cisa.gov/resources-tools/resources/ai-cybersecurity-collaboration-playbook">AI Cybersecurity Collaboration Playbook</a></li> <li><a href="https://www.cisa.gov/resources-tools/resources/secure-demand-priority-considerations-operational-technology-owners-and-operators-when-selecting">Secure by Demand: Priority Considerations for Operational Technology Owners and Operators when Selecting Digital Products</a></li> <li><a href="https://www.cisa.gov/resources-tools/resources/defending-against-software-supply-chain-attacks">Defending Against Software Supply Chain Attacks</a></li> <li><a href="https://www.cisa.gov/resources-tools/resources/2025-minimum-elements-software-bill-materials-sbom">2025 Minimum Elements for a Software Bill of Materials (SBOM)</a></li> <li><a href="https://www.cisa.gov/cybersecurity-performance-goals-2-0-cpg-2-0">Cybersecurity Performance Goals 2.0 (CPG 2.0)</a></li> </ul></div> <div> <h3>NSA resources</h3> <ul><li><a href="https://www.nsa.gov/Press-Room/Press-Releases-Statements/Press-Release-View/Article/4192332/nsas-aisc-releases-joint-guidance-on-the-risks-and-best-practices-in-ai-data-se/">AI Data Security: Best Practices for Securing Data Used to Train &amp; Operate AI Systems</a></li> <li><a href="https://www.nsa.gov/Press-Room/Press-Releases-Statements/Press-Release-View/Article/3741371/nsa-publishes-guidance-for-strengthening-ai-system-security/">Deploying AI Systems Securely: Best Practices for Deploying Secure and Resilient AI Systems</a></li> <li><a href="https://media.defense.gov/2024/Apr/15/2003439257/-1/-1/0/CSI-DEPLOYING-AI-SYSTEMS-SECURELY.PDF">Zero Trust Implementation Guideline Primer (PDF, 3.045 KB)</a></li> <li><a href="https://media.defense.gov/2026/Jan/08/2003852321/-1/-1/0/CTR_ZIG_DISCOVERY_PHASE.PDF">Zero Trust Implementation Guideline Discovery Phase (PDF, 2.124 KB)</a></li> </ul></div> <div> <h2>NCSC-UK resources</h2> <ul><li><a href="https://www.ncsc.gov.uk/collection/guidelines-secure-ai-system-development">Guidelines for secure AI system development </a></li> </ul></div> <div> <h2>Other resources</h2> <ul><li><a href="https://www.gov.uk/government/publications/ai-cyber-security-code-of-practice/code-of-practice-for-the-cyber-security-of-ai">UK Government: Code of Practice for the Cyber Security of AI</a></li> <li><a href="https://assets.publishing.service.gov.uk/media/679cae441d14e76535afb630/Implementation_Guide_for_the_AI_Cyber_Security_Code_of_Practice.pdf">UK Government: Implementation Guide for the AI Cyber Security Code of Practice (PDF, 989 KB)</a></li> <li><a href="https://www.etsi.org/committee/2312-sai">European Telecommunications Standards Institute: Technical Committee Securing Artificial Intelligence (TC SAI)</a></li> <li><a href="https://www.etsi.org/deliver/etsi_en/304200_304299/304223/02.01.01_60/en_304223v020101p.pdf">European Telecommunications Standards Institute: Securing Artificial Intelligence (SAI); Baseline Cyber Security Requirements for AI Models and Systems (PDF, 108 KB)</a></li> <li><a href="https://www.bsi.bund.de/SharedDocs/Downloads/EN/BSI/KI/SBOM-for-AI_Food-for-thoughts.html">G7 Cybersecurity Working Group: A Shared G7 Vision on Software Bill of Materials for Artificial Intelligence</a></li> <li><a href="https://www.nist.gov/itl/ai-risk-management-framework"><abbr title="National Institute of Standards and Technology">NIST</abbr>: AI Risk Management Framework</a></li> <li><a href="https://csrc.nist.gov/pubs/sp/800/207/final"><abbr title="National Institute of Standards and Technology">NIST</abbr>: SP 800-207 Zero Trust Architecture</a></li> <li><a href="https://genai.owasp.org/resource/owasp-top-10-for-agentic-applications-for-2026/"><abbr title="Open Web Application Security Project">OWASP</abbr>: Top 10 for Agentic Applications for 2026</a></li> </ul></div> </section><!–** TOP OF PAGE ******–><div class="pull-right small text-muted mrgn-bttm-0"><a href="#wb-tphp">Top of page</a> <span aria-hidden="true" class="text-primary glyphicon glyphicon-circle-arrow-up"></span></div> <!–** END TOP OF PAGE **–> <section><h2 id="App-a">Appendix A</h2> <p>Use the following cyber security prerequisites before implementation of AI agents.</p> <div> <h3>Design</h3> <ul><li>Implement strong authentication by following Secure by Design principles</li> <li>Design transparency requirements into the system architecture to enable detection of deception indicators</li> <li>Use frameworks, such as zero-trust authentication, the Application Security Verification Standard (ASVS), or OAuth2</li> <li>Build system infrastructure in a secure, sandboxed environment with encryption, rate limiting and sanitization</li> <li>Apply the principle of least privilege to assign only the minimal rights necessary for an agent’s task</li> <li>Limit entitlements to the exact resources, operations and timeframes needed</li> <li>Replace static, long-lived secrets with ephemeral credentials that expire when the job is complete</li> <li>Dynamically scope privilege for sub-tasks and revoke elevated rights immediately when finished, mitigating scope creep</li> <li>Build agent architecture with secure protocols and safe defaults that adhere to communication standards and security policies</li> <li>Implement message validation by default so that message components include integrity and freshness checks before use</li> </ul></div> <div> <h3>Development</h3> <ul><li>Use principles from the <a href="https://dodcio.defense.gov/Portals/0/Documents/Library/DoD%20Enterprise%20DevSecOps%20Fundamentals%20v2.5.pdf">U.S. Department of War Enterprise DevSecOps Fundamentals (PDF 2,811 KB) to secure the development</a></li> <li>Minimize agent and tool scope, and monitor components for unusual or unexpected behaviour</li> <li>Enforce tool understanding and only incorporate tools into systems that the owner fully understands and accepts the tool’s scope, including possible external effects and processes that it may trigger</li> <li>Refer to frameworks, such as the <a href="https://csrc.nist.gov/projects/ssdf"><abbr title="National Institute of Standards and Technology">NIST</abbr> Secure Software Development Framework</a> or <a href="https://slsa.dev/">Supply-chain Levels for Software Artifacts: Safeguarding artifact integrity across any software supply chain</a></li> <li>Use supply chain risk management practices for third-party dependencies</li> <li>Apply existing governance and organizational management policies</li> <li>Conduct threat modelling using frameworks, such as OWASP Top 10:2025, MITRE ATT&amp;CK and MITRE D3FEND and use the results to inform mitigations tailored to the operational environment</li> <li>Prepare incident response plans and team</li> </ul></div> </section></div> </div> </div> </div> </div> </article>

<article data-history-node-id="8183" about="/en/news-events/cyber-centre-urges-heightened-vigilance-amid-global-tensions-high-profile-events" class="cccs-basic-page full clearfix"> <div class="content"> <div class="layout layout–onecol"> <div class="layout__region layout__region–content"> <div data-block-plugin-id="extra_field_block:node:cccs_basic_page:links" class="block block-layout-builder block-extra-field-blocknodecccs-basic-pagelinks clearfix"> </div> <div data-block-plugin-id="field_block:node:cccs_basic_page:body" class="block block-layout-builder block-field-blocknodecccs-basic-pagebody clearfix"> <div class="field field–name-body field–type-text-with-summary field–label-hidden field–item"><p>The Communications Security Establishment Canada (CSE) and its Canadian Centre for Cyber Security (Cyber Centre) continue to urge Canadian organizations to remain vigilant during periods of heightened geopolitical tensions and high-profile events.</p> <p>Although malicious cyber activity can occur at any time, organizations should be especially vigilant during these periods as they can draw the attention of cyber threat actors, including ideologically motivated individuals or groups targeting organizations they perceive as adversaries. These actors may try to disrupt services, deface websites, attract publicity or undermine confidence in institutions and organizations.</p> <p>Threats may include distributed denial of service (DDoS) attacks against public-facing websites and online services, as well as website defacement and other disruptive activity. These attacks can disrupt operations, cause financial losses, damage reputations and reduce access to critical services. They may also be used to distract defenders from other malicious cyber activity targeting an organization at the same time.</p> <p>All Canadian organizations, especially essential service providers and critical infrastructure operators, should review their cyber security posture, and consult the following Cyber Centre resources.</p> <ul><li>Read our previous <a href="/en/news-events/cyber-centre-statement-distributed-denial-service-campaign-targeting-multiple-canadian-sectors">statement on a distributed denial of service (DDoS) campaign targeting multiple Canadian sectors</a></li> <li>Adopt our <a href="/en/cyber-security-readiness/cross-sector-cyber-security-readiness-goals-toolkit">Cross-Sector Cyber Security Readiness Goals</a></li> <li>Consult our <a href="/en/guidance/top-10-it-security-actions-protect-internet-connected-networks-and-information-itsm10089">Top 10 IT security actions to protect Internet connected networks and information</a></li> <li>Review and implement the advice and guidance on: <ul><li><a href="/en/guidance/defending-against-distributed-denial-service-ddos-attacks-itsm80110">Defending against distributed denial of service (DDoS) attacks</a></li> <li><a href="/en/guidance/distributed-denial-service-attacks-prevention-and-preparation-itsap80110">Distributed denial of service attacks – prevention and preparation</a></li> <li><a href="/en/alerts-advisories/distributed-denial-service-campaign-targeting-multiple-canadian-sectors">Distributed Denial of Service campaign targeting multiple Canadian sectors</a></li> <li><a href="/en/guidance/website-defacement-itsap00060">Website defacement</a></li> <li><a href="/en/guidance/developing-your-incident-response-plan-itsap40003">Developing your incident response plan</a></li> <li><a href="/en/guidance/security-considerations-critical-infrastructure-itsap10100">Security considerations for critical infrastructure</a></li> <li><a href="/en/guidance/protect-your-operational-technology-itsap00051">Protect your operational technology</a></li> <li><a href="/en/news-events/joint-guidance-principles-operational-technology-cyber-security">Joint guidance on the principles of operational technology cyber security</a></li> <li><a href="/en/news-events/joint-guidance-isolating-vital-systems">Joint guidance on isolating vital systems</a></li> </ul></li> <li>Explore our <a href="/en/cyber-security-readiness/critical-infrastructure-resilience-escalated-threat-navigation-initiative">Critical infrastructure resilience and escalated threat navigation initiative</a></li> </ul><p>The Cyber Centre continues to work closely with government, critical infrastructure operators, industry partners, and international allies to monitor the cyber threat environment and support the protection of Canadian systems and networks.</p> <p>Organizations should act now to strengthen their cyber defences and promptly report cyber incidents or suspicious cyber activity to the Cyber Centre through our <a href="https://www.cyber.gc.ca/en/incident-management">online portal</a>.</p> </div> </div> </div> </div> </div> </article>

<article data-history-node-id="8160" about="/en/news-events/g7-cybersecurity-working-group-call-action-preparing-post-quantum-era" class="cccs-basic-page full clearfix"> <div class="content"> <div class="layout layout–onecol"> <div class="layout__region layout__region–content"> <div data-block-plugin-id="extra_field_block:node:cccs_basic_page:links" class="block block-layout-builder block-extra-field-blocknodecccs-basic-pagelinks clearfix"> </div> <div data-block-plugin-id="field_block:node:cccs_basic_page:body" class="block block-layout-builder block-field-blocknodecccs-basic-pagebody clearfix"> <div class="field field–name-body field–type-text-with-summary field–label-hidden field–item"><p>Canada is joining its G7 Cybersecurity Working Group partners in releasing the <strong>Preparing for the Post-Quantum Era: A Call to Action</strong>, a joint publication highlighting the growing risks that quantum computing poses to current public-key cryptography and urging organizations to begin transitioning to post-quantum cryptography (PQC) now.</p> <p>The publication emphasizes that while the timeline for cryptographically relevant quantum computers (CRQCs) remains uncertain, organizations face immediate risks such as "harvest now, decrypt later" attacks. In these attacks, threat actors collect encrypted data now, with the intention of decrypting it in the future using quantum computing capabilities. It also warns that quantum-enabled attacks could compromise authentication mechanisms, enabling attackers to impersonate trusted entities, forge data, and undermine secure communications.</p> <p>The publication recommends ways to support a coordinated global transition to <abbr title="post-quantum cryptography">PQC</abbr>. To help organizations and governments prepare for the transition, the publication outlines five priority areas for action:</p> <ul><li>raising awareness of quantum-related cybersecurity risks and the benefits of <abbr title="post-quantum cryptography">PQC</abbr></li> <li>developing national strategies to support the adoption and availability of <abbr title="post-quantum cryptography">PQC</abbr> technologies</li> <li>advancing research and development to support innovation and practical deployment of <abbr title="post-quantum cryptography">PQC</abbr> solutions</li> <li>strengthening public-private partnerships to share expertise, resources, and best practices</li> <li>integrating <abbr title="post-quantum cryptography">PQC</abbr> into cybersecurity requirements</li> </ul><p>The publication calls on public and private sector organizations to adopt a phased, risk-based approach to <abbr title="post-quantum cryptography">PQC</abbr> migration. This includes taking inventory of cryptographic assets, identifying critical systems, mapping dependencies, and developing transition plans to help protect sensitive data and digital infrastructure.</p> <p>Consult the full publication: <a href="https://cyber.gouv.fr/en/publications/jointly-led-international-publications/preparing-for-the-post-quantum-era-a-call-to-action/">Preparing for the post-quantum era: a call to action</a></p> <h2>Learn more</h2> <ul><li><a href="/en/news-events/g7-cybersecurity-working-group-statement-preparing-post-quantum-cryptography-migration">G7 Cybersecurity Working Group statement on preparing for a post-quantum cryptography migration</a></li> <li><a href="/en/guidance/preparing-your-organization-quantum-threat-cryptography-itsap00017">Preparing your organization for the quantum threat to cryptography (ITSAP.00.017) – Canadian Centre for Cyber Security</a></li> <li><a href="/en/guidance/roadmap-migration-post-quantum-cryptography-government-canada-itsm40001">Roadmap for the migration to post-quantum cryptography for the Government of Canada (ITSM.40.001)</a></li> </ul></div> </div> </div> </div> </div> </article>

This joint guidance explains why effective communications during outages is critical to minimize operational impacts, maintain credibility and situational awareness, and support response efforts.

<article data-history-node-id="8082" about="/en/news-events/joint-guidance-opportunities-artificial-intelligence-cyber-defence" class="cccs-basic-page full clearfix"> <div class="content"> <div class="layout layout–onecol"> <div class="layout__region layout__region–content"> <div data-block-plugin-id="extra_field_block:node:cccs_basic_page:links" class="block block-layout-builder block-extra-field-blocknodecccs-basic-pagelinks clearfix"> </div> <div data-block-plugin-id="field_block:node:cccs_basic_page:body" class="block block-layout-builder block-field-blocknodecccs-basic-pagebody clearfix"> <div class="field field–name-body field–type-text-with-summary field–label-hidden field–item"><p>The Canadian Centre for Cyber Security (Cyber Centre) has joined the Australian Signals Directorate’s Australian Cyber Security Centre (ASD’s ACSC) and the following international partners in releasing cyber security guidance on opportunities for artificial intelligence (AI) in cyber defence:</p> <ul><li>New Zealand’s National Cyber Security Centre (NCSC-NZ)</li> <li>United Kingdom’s National Cyber Security Centre (NCSC-UK)</li> </ul><p>AI presents a significant opportunity for cyber defenders. When used safely, securely and responsibly, AI can:</p> <ul><li>strengthen prioritisation of cyber risks</li> <li>improve detection of threats and vulnerabilities</li> <li>support faster response and recovery</li> <li>reduce reliance on repetitive manual tasks</li> </ul><p>The joint guidance explains how organizations can use AI to improve their cyber security while also managing the risks that come with using AI. The guidance describes how the cyber security environment is changing and shows how AI can support key security activities, including governance, identifying risks, protecting systems, detecting threats, responding to incidents, and recovering from cyber attacks.</p> <p>The guidance also provides practical principles for adopting AI securely and includes important questions that cyber security teams should ask AI vendors to help ensure AI tools are used safely and securely.</p> <p>Consult the full joint guidance: <a href="https://www.cyber.gov.au/business-government/secure-design/artificial-intelligence/opportunities-for-ai-in-cyber-defence?ref=search">Opportunities for AI in cyber defence – Use of AI by cyber-security teams</a></p> </div> </div> </div> </div> </div> </article>

<article data-history-node-id="8032" about="/en/guidance/security-considerations-electronic-vote-tabulators-itsm10102" class="cccs-basic-page full clearfix"> <div class="content"> <div class="layout layout–onecol"> <div class="layout__region layout__region–content"> <div data-block-plugin-id="extra_field_block:node:cccs_basic_page:links" class="block block-layout-builder block-extra-field-blocknodecccs-basic-pagelinks clearfix"> </div> <div data-block-plugin-id="field_block:node:cccs_basic_page:body" class="block block-layout-builder block-field-blocknodecccs-basic-pagebody clearfix"> <div class="field field–name-body field–type-text-with-summary field–label-hidden field–item"><div class="row"> <div class="col-md-4 pull-left hidden-xs hidden-sm"> <p class="text-left"><strong>August 2026</strong></p> </div> <div class="col-md-4 hidden-xs hidden-sm"> <p class="text-center"><strong>Management series</strong></p> </div> <div class="col-md-4 pull-right hidden-xs hidden-sm"> <p class="text-right"><strong>ITSM.10.102</strong></p> </div> <!–MOBILE STARTS HERE–> <div class="hidden-lg hidden-md text-center"> <p><strong>August 2026 | Management series</strong></p> </div> </div> <p>The use of ballot counting systems, also known as vote tabulation technology, to replace manual ballot counting during democratic elections is gaining wider adoption. These technologies can help improve the efficiency of the ballot tabulation process, as well as reduce human-related errors during the ballot counting process.</p> <p>This publication provides guidance on the cyber security considerations and system requirements for deploying vote tabulators during democratic elections. It highlights the key risks, threats and safeguards that elections management authorities should assess when considering ballot counting systems.</p> <section><details class="mrgn-tp-md"><summary><h2 class="h3">Table of contents</h2> </summary><ul class="list-unstyled"><li><a href="#1">Scope</a></li> <li><a href="#2">Electronic voting</a> <ul><li><a href="#fig1">Figure 1: Generic elections system architecture</a></li> </ul></li> <li><a href="#3">Ballot counting and tabulation</a> <ul><li><a href="#3.1">Types of vote tabulators</a></li> </ul></li> <li><a href="#4">Threat analysis</a></li> <li><a href="#5">Security control considerations</a> <ul><li><a href="#5.1">Physical and hardware security</a></li> <li><a href="#5.2">Lifecycle security</a></li> <li><a href="#5.3">Continuous risk assessment</a></li> <li><a href="#5.4">Protect your data</a></li> <li><a href="#5.5">Network isolation safeguards</a></li> <li><a href="#5.6">Secure access controls</a></li> <li><a href="#5.7">Validate and authenticate devices</a></li> <li><a href="#5.8">Patches and system updates</a></li> <li><a href="#5.9">Secure application development</a></li> <li><a href="#5.10">Audit log monitoring</a></li> <li><a href="#5.11">Business continuity</a></li> <li><a href="#5.12">Educate users</a></li> <li><a href="#5.13">Security control effectiveness</a></li> </ul></li> <li><a href="#6">Conclusion</a></li> </ul></details></section><section class="mrgn-tp-md"><p>Adopting digital technologies to improve the efficient delivery of the electoral process is an increasingly attractive proposition for modern democratic elections. Elections administrators are embracing the use of modern digital methods to improve or replace traditional and manual voting processes. Manual counting can be subject to human errors from fatigue, eyestrain, or distractions, which can lead to ballots being misjudged and misallocated. In very tight electoral races, those errors can have serious consequences. Using electronic vote counting technologies to support aspects of the elections process can help prevent human-related computational errors, expand voter accessibility options, and improve election transparency. These technologies can also allow for the timely reporting of elections results.</p> <p>Although federal elections in Canada do not use vote tabulators or ballot counting systems, these systems have been used in provincial and municipal government elections. Vote tabulators are electronic devices used to automate the counting of marked paper ballots. Some specific models may support the storage of digital ballot records, integration with assistive devices or vote capturing capabilities, and other additional functionalities. However, despite their potential benefits, the associated risks must be carefully considered.</p> <p>Replacing traditional processes with vote tabulators can have many advantages; however, this can introduce new vulnerabilities and threats into the democratic process, allowing threat actors opportunities to inflict harm. These threats may come from sophisticated nation-state actors or unprincipled stakeholders interested in disrupting the peaceful conduct of the democratic process. Before deploying vote tabulators, elections authorities should conduct a detailed risk assessment to evaluate the potential risks and associated security threats. The Cyber Centre’s <a href="/en/guidance/cyber-threats-canadas-democratic-process-2025-update">Cyber threats to democratic process: 2025 update</a> assesses that the likelihood for state-sponsored cyber actors to target Canadian elections is <strong>almost certain</strong>. It is important to understand specific risks and threats that may be associated with electronic vote tabulation deployments.</p> <p>In this publication, we discuss important cyber security considerations for elections administrators when deciding whether to use vote tabulators, the recommended secure architecture model for deploying these devices, as well as the security controls required to safeguard the integrity of the system. Note that it is difficult to establish suitably secure network connectivity for any election’s infrastructure.</p> </section><h2 id="1">Scope</h2> <p>This guidance applies to vote-tabulation systems that process paper ballots only, not systems that handle electronic ballots. While some recommendations may also apply to electronic ballot systems, they may not fully address the distinct threats those systems face.</p> <h2 id="2">Electronic voting</h2> <p>Electronic voting system architectures are complex and can involve many disparate systems. An electronic voting system architecture describes the information systems components involved in electronic voting and how these components interact with each other. These systems may include:</p> <ul><li>public websites</li> <li>party registration systems</li> <li>voter registrations systems</li> <li>ballot creation and printing systems</li> <li>election management systems</li> <li>online voting systems</li> <li>ballot counting or vote tabulation systems</li> <li>results publishing systems</li> </ul><p>These systems often run distinct software and firmware technologies, can be interconnected, and may communicate across diverse network protocols. Figure 1 depicts a generic modern electronic voting systems architecture, based on the Center for Internet Security’s (CIS) <a href="https://www.google.com/url?sa=t&amp;rct=j&amp;q=&amp;esrc=s&amp;source=web&amp;cd=&amp;ved=2ahUKEwiyvL6B0MiCAxXQJjQIHXDTB0wQFnoECBYQAQ&amp;url=https%3A%2F%2Fwww.cisecurity.org%2Fwp-content%2Fuploads%2F2018%2F02%2FCIS-Elections-eBook-15-Feb.pdf&amp;usg=AOvVaw0Dq56eE-JFpTwo7r7o9ENs&amp;opi=89978449">Handbook for Elections Infrastructure Security</a>.</p> <div class="panel panel-default"> <div class="panel-body"> <figure id="fig1"><figcaption class="text-center"><strong>Figure 1: Generic elections system architecture</strong></figcaption><img alt="Generic elections system – Long description immediately follows" class="img-responsive" src="/sites/default/files/images/itsm.10.102-fig1-e.png" /></figure><p><strong>Figure 1: Generic elections systems architecture</strong> presents a visual representation of a generic electronic voting system architecture. The image provides additional details on how ballots go from creation and printing to having the results counted and tabulated to the results being published.</p> <p>The components outlined in the image represent the entirety of a generic elections system architecture, including:</p> <ol><li>voter registration systems</li> <li>party and candidate registration and verification</li> <li>campaign finance management</li> <li>election management systems</li> <li>ballot creation and printing</li> <li>electronic poll books</li> <li>ballot capturing devices</li> <li>ballot counting and tabulation <ul><li>ballot scanners</li> <li>ballot counting and tabulators</li> </ul></li> <li>results transmission and publishing</li> </ol></div> </div> <p>This guidance is focused on the ballot counting and ballot tabulation components of this architecture. Our recommendations address core cyber security and operational issues that require attention to safely operate and deploy these systems. First, we provide a generic description for ballot counting and tabulation systems.</p> <!–** TOP OF PAGE ******–> <div class="clearfix"> </div> <div class="pull-right small text-muted mrgn-bttm-0"><a href="#wb-tphp">Top of page</a> <span aria-hidden="true" class="text-primary glyphicon glyphicon-circle-arrow-up"></span></div> <!–** END TOP OF PAGE **–> <h2 id="3">Ballot counting and tabulation</h2> <p>Ballot counting and tabulation is the process of counting and aggregating used and unused ballots in an election. The process involves gathering valid and invalid (spoiled) ballots, counting unused ballots, and aggregating valid ballots for each candidate in the election. Ballot counting and tabulation can be completed manually (namely by hand-counting) or by using electronic systems such as vote tabulators. Both methods can introduce errors in the process for various reasons. For example, if elections workers are tired or distracted, counting ballots by hand may lead to computational errors. Similarly, software flaws or compromises of electronic vote tabulators may lead to votes being inaccurately categorized. In general, it is important for elections administrators to continually assess counting methods and implement mechanisms to determine or validate that their counting processes are operating within accepted error rates.</p> <p>The ballot counting method used is often determined by legal provisions. For example, the <a href="https://laws-lois.justice.gc.ca/eng/acts/E-2.01/index.html"><em>Canada Elections Act</em></a> requires ballots cast during federal elections to be hand-counted, while some provinces, territories and municipalities have changed local laws to allow the use of electronic vote tabulators in their local elections.</p> <p>At a high level, a typical ballot counting and tabulation process includes the following activities:</p> <ul><li>validating each ballot</li> <li>sorting and counting valid or invalid (spoiled) cast ballots</li> <li>aggregating valid cast ballots for each candidate</li> <li>generating a tabulated summary from tallied ballots</li> </ul><p>Note that the transmission of election outcomes is considered a separate process, as this may require additional system components for transmitting and sharing that information. Results aggregation may also require additional collation at a central facility before official results are released.</p> <p>The following are some methods that may be used to tabulate ballots during elections.</p> <h3 class="h4">Manual counting and tabulation</h3> <p>In this method, elections officials manually sort, validate, and count casted paper ballots. All activities are done by hand, including generating election results tables. As noted by Elections Canada in the <a href="https://www.elections.ca/content.aspx?section=res&amp;dir=pub/ecdocs/rom/vIII/ch_2&amp;document=ch_2&amp;lang=e">Returning Officer’s Manual</a>, in Canadian federal elections, ballots are counted manually at polling stations, and preliminary results are then transmitted (often by telephone) to Elections Canada .</p> <p>For elections where a different tabulation method is primarily used (e.g. provincial or municipal elections), manual ballot counting and tabulation are commonly used as a fallback measure. Manual ballot counting and tabulation are also used to perform election audits. Manually counting and tabulating ballots can be tedious and time-consuming, especially for polling stations with higher voter density. While manual counting processes may provide a greater degree of trust due to their inherent simplicity and easier security auditability, they may allow for human subjectivity and errors.</p> <h3 class="h4">Electronic counting and tabulation</h3> <p>This method involves using electronic vote tabulators to sort, validate, and count election ballots. This method can be used to count both paper and digital ballots. Paper ballots are fed into and analyzed by the device while digital ballots are captured and processed. Electronic vote tabulators may have built-in printing capabilities or may support connectivity with external printers. Note that any external or peripheral devices connected to a vote tabulator must adhere to the same security control standards as the vote tabulator system. We recommend maintaining independence between the counting system and the ballot creation system so that a threat or attack against one does not impact the other.</p> <h3 id="3.1">Types of vote tabulators</h3> <p>There are different types of vote tabulators, often defined by their function, capabilities, and use cases.</p> <h4>Unit vote tabulators</h4> <p>These are digital scan machines that read and record ballots at a polling station. They are designed for light use only. They may or may not contain additional programmable capabilities which can assist with ballot validation, sorting, calculations, and local results processing.</p> <h4>Central tabulators</h4> <p>These are devices used to process ballots at a central location. They are typically industrial-sized and designed to process large amounts of ballots. They can also be used to rapidly perform large-scale recounts or to process mail-in ballots.</p> <h4>Optical scan tabulators</h4> <p>These are optical scan readers commonly used to help process marked paper ballots. These scanners rely on optical character recognition (OCR) or optical mark recognition (OMR) technology to read ballots. The scanner generates a digital representation of the original ballot which can then be used for additional processing. Optical scan systems are often used to process mail-in ballots.</p> <h2 id="4">Threat analysis</h2> <p>The Cyber Centre periodically releases an assessment of <a href="/en/guidance/cyber-threats-canadas-democratic-process-2025-update">Cyber threats to Canada’s democratic process (TDP)</a>. In its most update, the Cyber Centre assesses that democratic processes across the globe remain a target of interest for cyber threat actors and that the likelihood for state-sponsored cyber actors to target Canadian elections is <strong>almost certain</strong>. Vote tabulation systems may certainly introduce new vulnerabilities within the electoral process and may become a target for motivated, state-sponsored threat actors. Election agencies should conduct detailed threat and risk assessments to ensure that they understand the risks associated with using these systems and that vote tabulator systems are protected against such threats.</p> <p>Threat actors often seek to undermine public confidence in the legitimacy of the democratic process. According to the 2025 <abbr title="Cyber threats to Canada’s democratic process">TDP</abbr> report, state-sponsored actors may attempt to weaken trust by altering content on devices used during the elections process. Threat actors can infiltrate and compromise the supply chains used to procure vote tabulator devices. They can also launch disruptive attacks against these devices to render them inoperative, thereby impacting the smooth conduct of elections. Malicious actors can use targeted misinformation or disinformation campaigns to spread inaccuracies about the operation of vote tabulator devices, thereby undermining voter trust and confidence.</p> <p>State-sponsored threat actors have conducted numerous attacks against electoral systems and processes in recent years. For example, Ukraine, Estonia, Ecuador, and even the <abbr title="United Kingdom">UK</abbr>, have been targeted. Some of the attack vectors have included deploying disruptive malware on servers used to tabulate votes and planting fake results with the intention of releasing them publicly.</p> <p>Based on the prevailing threat environment, it is very likely that cybercriminals and state-sponsored actors will target ballot counting and tabulation processes to achieve their overall objectives. The following are ways threat actors could launch attacks and target ballot counting and tabulation systems:</p> <ul><li>taking over vote tabulation systems for the purpose of altering the results</li> <li>infiltrating device or systems-component manufacturers to compromise the supply chain for ballot counting and tabulation devices</li> <li>compromising elections record databases, resulting in unauthorized disclosure of confidential electronic voter records and undermining confidence in the democratic process</li> <li>disrupting vote tabulation system functions through exhaustion and destruction attacks such as distributed denial of service (DDoS) and ransomware attacks</li> <li>impersonating elections authorities to spread misinformation or falsehoods about the integrity of the process or devices to sow mistrust and erode confidence in the democratic process</li> <li>compromise privileged credentials to gain access to back-end systems to alter tabulation records or trigger further disruptive attacks</li> </ul><!–** TOP OF PAGE ******–><div class="clearfix"> </div> <div class="pull-right small text-muted mrgn-bttm-0"><a href="#wb-tphp">Top of page</a> <span aria-hidden="true" class="text-primary glyphicon glyphicon-circle-arrow-up"></span></div> <!–** END TOP OF PAGE **–> <h2 id="5">Security control considerations</h2> <p>In this section, we highlight important security controls you should consider when acquiring, deploying, and operating vote tabulation systems. These recommendations are aligned with the Cyber Centre’s <a href="/en/guidance/cyber-security-privacy-risk-management/itsp10033">Cyber security and privacy risk management: A lifecycle approach (ITSP.10.033)</a>. Election administrators should consider foundational system design and security principles to safeguard hardware, firmware, software components, and related processes.</p> <h3 id="5.1">Physical and hardware security</h3> <p>Physical layer protections in vote tabulator systems are critical in ensuring that other security controls can be trusted, and that the entire system operates as intended. Vote tabulators must be designed, manufactured, operated, and decommissioned in accordance with secure by design principles to better protect them against threats, such as supply chain compromise during any stage of its lifecycle.</p> <h4>Design systems to withstand physical and environmental hazards</h4> <p>Vote tabulators should be designed for operational resilience . The devices, storage media, and other system components should be able to withstand physical and environmental hazards, such as physical shocks, high humidity, and temperature fluctuations, to protect against data loss. These devices should also have an emergency power supply; an intelligent back-up power system will also allow for graceful power-down processes. It is important to protect devices and their electronic data from power surges and outages. We also recommend having electromagnetic shielding and tamper-evident seals as feature considerations.</p> <h4>Manufactured for purpose</h4> <p>Vote tabulators should be manufactured for a single purpose to ensure security functions are operating within defined security boundaries. Single purpose systems avoid the complexity associated with multi-use designs, while also ensuring system functions are strictly defined and security restrictions can be adequately enforced. General-purpose or multi-use systems extend the attack surface, providing additional opportunities for the adversary to compromise. We recommend that vote tabulators should be purpose-built and dedicated, single-use systems.</p> <h4>Operate systems with strict physical access controls</h4> <p>When deploying and operating vote tabulators, you should implement strict physical access controls. Only authorized individuals on approved access lists should have physical access to the devices before, during, and after elections. The locations where these devices are stored, deployed, or transported throughout their lifecycle should be monitored, and mechanisms should be in place to detect and deter unauthorized access. When choosing where to deploy vote tabulators, you should consider how you will secure them from unauthorized observation and access. Device screens should be set up to prevent on-screen information from being read or observed from a distance. Electoral management bodies (EMBs) should maintain access records for each device throughout its lifecycle, from acquisition to decommissioning.</p> <h4>Monitor systems and maintain strong chain of custody records</h4> <p>Maintaining chain of custody records plays a significant role in ensuring the physical and logical integrity of the vote tabulators. Elections administrators should implement rigorous chain of custody practices to prevent tampering and other malicious device alterations. The following are some ways of protecting vote tabulators and the data contained therein.</p> <ul><li>Record all vote tabulator devices movement and access throughout their lifecycles</li> <li>Use tamper-evident seals</li> <li>Transport vote tabulators in secure containers to and from polling centres</li> <li>Document and verify chain of custody when transporting vote tabulators</li> <li>Protect and authenticate all physical access to vote tabulation equipment</li> <li>Review security monitoring controls before, during, and after elections to validate their effectiveness</li> <li>Decommission devices when chain of custody records cannot be securely validated</li> </ul><p>Device decommissioning procedures must ensure devices are cryptographically sanitized before disposal or destruction. Use only approved sanitization methods appropriate for the device media and use-case. Some sanitization methods are not effective or suitable on some media types. For example, solid-state drives (SSD) require at least a double overwrite pass and a secure erasure (SE) to achieve the sanitization objectives. For more information on Cyber Centre recommended guidance, refer to <a href="/en/guidance/it-media-sanitization-itsp40006">IT media sanitization (ITSP.40.006)</a>.</p> <h3 id="5.2">Lifecycle security</h3> <p>Democratic processes may be vulnerable to planned, persistent threats because of system lifecycle risks. In the context of information technology (IT) security risk management, “lifecycle” refers to the design/development, acquisition, integration/installation, operation, monitoring, maintenance, and disposal of <abbr title="information technology">IT</abbr> assets which, for the purposes of this guidance, include electronic vote tabulators. Examples of lifecycle risks to electronic vote tabulators include procuring them through unsecure supply chains or having them designed or developed by unverified or untrusted manufacturers. Therefore, you should keep security in mind at every stage of a vote tabulator’s lifecycle and address and mitigate potential risks that could allow threat actors to compromise the integrity of an election.</p> <h4>Acquire devices through trusted supply chains</h4> <p>Vote tabulation systems typically contain millions of individual micro-components, as well as component sub-systems such as an image scanner unit, a paper ballot verification unit, device firmware or software, power systems, data storage units, and others. Threat actors can infiltrate supply chains to compromise the security of these devices and systems. The complexity of modern system lifecycles makes it difficult for commercially available systems to guarantee protection against supply chain embedded compromises. As a result, organizations should assume such compromises may exist on their device. We recommend adopting strategies to assess and mitigate supply chain-related risks by following the <a href="/en/guidance/cyber-supply-chain-approach-assessing-risk-itsap10070">Cyber Centre’s approach to assessing cyber supply chain risks</a>.</p> <p>Elections officials should source vote tabulators and system components through trusted supply chains and physically inspect the devices before and after acquiring them, as well as before, during and after elections. Vote tabulators should have tamper-evident seals, tape, or security labelling to aid the detection of unauthorized physical access or unauthorized changes to hardware or physical system components.</p> <p>Elections administrators should establish secure procurement procedures to limit the potential for threat actors to infiltrate supply chains. The following are some actions that could ensure the security of the vote tabulator infrastructure supply chain:</p> <ul><li>Establish a supply chain risk management (SCRM) program to assess risks associated with procurement processes <ul><li>Assess and vet all technology suppliers to ensure underlying processes and relationships are secure</li> <li>Periodically assess suppliers and their cyber security practices</li> </ul></li> <li>Ensure that only trusted and vetted supplier personnel have access to sensitive electoral and system data throughout the system development lifecycle</li> <li>Collaborate with technology vendors to ensure verification processes are in place <ul><li>Ensure that the components of vote tabulators are genuine and have not been compromised through third-party relationships</li> <li>Specifically, work with vote tabulator manufacturers to implement tools to validate the integrity of hardware, firmware, and software running on tabulation devices</li> </ul></li> <li>Election agencies can also request information on detailed Hardware Bill of Materials (HBOM) or Software Bill of Materials (SBOM) to adequately track vulnerability risks associated with hardware and software components</li> </ul><p>For more information on supply chain cyber security, read <a href="/en/guidance/supply-chain-threats-and-commercial-espionage">Supply chain threats and commercial espionage</a> and <a href="/en/guidance/supply-chain-security-small-and-medium-sized-organizations-itsap00070">Supply chain security for small and medium-sized organizations (ITSAP.00.070)</a>.</p> <h3 id="5.3">Continuous risk assessment</h3> <p>Electoral management authorities should conduct a comprehensive risk analysis of the use of vote tabulators and understand the potential impact on the security of the entire electoral process. Assessing the risks of using vote tabulators should not be done in isolation.</p> <h4>Ascertain the risks associated with using vote tabulation systems</h4> <p><abbr title="electoral management bodies">EMBs</abbr> should consider technology in elections from a risk-management approach. While vote tabulators may offer opportunities for improving elections operations, it is important to understand the associated risks, which then may be mitigated, accepted, or avoided. Transferring the risk is not acceptable, as a compromise of a democratic process can have substantial reputational impacts. <abbr title="electoral management bodies">EMBs</abbr> should also avoid using vote tabulators if they are unprepared or unwilling to accept the associated risks. For more information on evaluating and managing risk, please consult our guidance <a href="/en/guidance/cyber-security-privacy-risk-management/itsp10033">Cyber security and privacy risk management: A lifecycle approach</a> and Security risk assessment for online voting systems supporting democratic processes (ITSM.10.103)</p> <h3 id="5.4">Protect your data</h3> <p>Electronic vote tabulators process, create, and store ballot data. Data is a critical asset for any business process, and especially for democratic processes that rely on digital systems. Therefore, you should have security mechanisms in place to protect the confidentiality, integrity, and availability of all forms of data, whether it is stored on the tabulation device or is in transit for processing.</p> <h4>Establish a holistic data security strategy</h4> <p>A holistic election data security strategy should be established to govern and secure electoral data throughout its lifecycle. This strategy should include data policies, acceptable handling procedures, and guidelines for managing data on vote tabulators. Data security policies may include recommendations on storage media types, secure data processing mechanisms, cryptographic standards, storage data formats, and data encryption. You may also consider performing a data security assessment to identify all data flows, paths, inputs, and outputs to and from the vote tabulator. However, there is no fully secure way to connect vote tabulators to electronic networks. Your data security strategy should address how to secure vote tabulator data throughout its lifecycle, from creation to destruction. For example, it should address how you plan to secure physical and digital ballots, restrict and manage access, enforce data encryption, and ensure compliance with legal requirements wherever your data may reside.</p> <h4>Appropriately classify data</h4> <p>Data classification represents a crucial component upon which many other security activities rely, especially when identifying appropriate security control mechanisms to protect the data. Data on vote tabulators typically exists at different classification levels and the choice of safeguards or security controls should take this into account. The vote tabulation system should be designed based on the highest classification of data or system components. For example, vote tabulators may be required to process cryptographic keys and data validation certificates may be used to encrypt and validate the integrity of election data.</p> <p>Cryptographic keys represent highly sensitive data and may necessitate higher data classification. The vote tabulation system must be designed to implement security controls at the highest level determined. For more information on data classification assessments, read our publication <a href="/en/guidance/cyber-security-privacy-risk-management/itsp10033">Cyber security and privacy risk management: Security and privacy controls and assurance activities catalogue (ITSP.10.033)</a>.</p> <p>Classifying data properly, together with implementing matching secure handling controls, can reduce the risk of a data compromise. Elections authorities should put processes in place to ensure that assets are only accessible to users on a ‘need to know’ basis and that electoral systems are only accessed using devices with the appropriate security authorization. Furthermore, only authorized devices should be used for processing and disseminating of elections-related data.</p> <h4>Protect data with cryptographic controls</h4> <p>Vote tabulator data should be protected with cryptographic mechanisms whether it is at rest or in transit (including on portable devices). Ballot data and vote counts should be secured against unauthorized disclosure and integrity attacks prior before and after they are officially released through authorized channels. Encryption can help protect data from unauthorized disclosure and modification. You should identify where the data resides at various points and then ensure you apply appropriate security controls to mitigate the risk of unauthorized use or disclosure. We recommend using only approved cryptographic algorithms and protocols. For more information on recommended cryptographic algorithms, read <a href="/en/guidance/cryptographic-algorithms-unclassified-protected-protected-b-information-itsp40111">Guidance on cryptographic algorithms for Unclassified, Protected A, and Protected B information (ITSP.40.111</a>). Maintaining a secure cryptographic key management system is important. Ensure you securely manage when, how, and where your encryption keys are created, stored, and destroyed. Implement a secure backup procedure for encryption keys to enable data recovery.</p> <h4>Consider data privacy protections</h4> <p>In Canada, election data is subject to privacy laws, including the <a href="https://laws-lois.justice.gc.ca/eng/acts/P-21/"><em>Privacy Act</em></a>. These laws mandate that a minimum set of security controls, such as data encryption and access management controls, be implemented to protect information about the voting public and ballots. We recommend you implement appropriate data privacy protections for ballots and voters’ information. The <a href="https://laws-lois.justice.gc.ca/eng/acts/E-2.01/index.html"><em>Canada Elections Act</em></a> requires Elections Canada to share voting data, including elections results, with political parties and provincial governments. Vote tabulators must be evaluated to ascertain compliance and alignment with ballot secrecy and privacy laws. Data residency considerations should also be evaluated, as election laws may require data (in-transit or at rest) to reside within specific geographical boundaries.</p> <!–** TOP OF PAGE ******–> <div class="clearfix"> </div> <div class="pull-right small text-muted mrgn-bttm-0"><a href="#wb-tphp">Top of page</a> <span aria-hidden="true" class="text-primary glyphicon glyphicon-circle-arrow-up"></span></div> <!–** END TOP OF PAGE **–> <h3 id="5.5">Network isolation safeguards</h3> <p>Due to the threat environment and the associated risks posed by state-sponsored threat activity, we recommend that you do not connect vote tabulators to any network at any time. Prior to elections, vote tabulators should be configured without any network connectivity, including Bluetooth and Wi-Fi, and should never be connected to the Internet at any point of their lifecycle. Data should be transferred to and from vote tabulators using non-network procedures, such as trusted portable storage devices or paper-based procedures.</p> <p>Strict network isolation (airgaps) is the most effective means of mitigating pre-existing compromises that may have been introduced through lifecycle attacks. Network isolation greatly increases the amount of effort required to gain remote command and control of a compromised system and forces an adversary to activate or control the compromised vote tabulator or related system through an alternative data channel, such as acoustic, or through procedural means, such using a compromised systems administrator. Certain serious attacks, like “logic bombs” that autonomously activate based on pre-set conditions, could remain on the system even without network connectivity, but the scope and sophistication of the possible attacks is much smaller without network command and control.</p> <h4>Establish strict network defences</h4> <p>Although deploying vote tabulators on a network is not recommended, elections administrators may still choose to deploy these systems on a network. In such cases, the network infrastructure supporting vote tabulators should be protected against interception, unauthorized data disclosure, modification, and denial of service attacks. You should implement measures to protect against ransomware attacks that may target the tabulation systems. You should track vulnerabilities and fully patch all devices before deploying them for elections. The following are some actions that you can take to implement a defence-in-depth or layered defensive approach to protecting your networks from threat actors.</p> <ul><li>Use firewalls or network zoning protections to restrict network access to sensitive assets on the network</li> <li>Consider deploying network threat detection/protection systems to mitigate network-based attacks</li> <li>Update network devices regularly to minimize exposure to vulnerabilities</li> <li>Harden network devices by disabling unnecessary services, ports, protocols and applications</li> <li>Avoid connecting vote tabulators to public or untrusted networks</li> <li>Disable Internet connectivity on vote tabulators or interconnected systems</li> <li>Implement protection against threats associated with third-party (vendors or partners) network connectivity</li> </ul><h3 id="5.6">Secure access controls</h3> <p>When designing and building an electronic tabulation system, it is essential to properly manage user and administrative privileges, isolate sensitive functions, and incorporate security engineering principles. User access rights should be restricted to only those required to perform assigned tasks. System functions should implement the principle of least privilege, which only gives users the set of privileges that is essential for them to perform authorized tasks. The following are some recommended actions you can take to implement secure access controls within vote tabulation systems.</p> <ul><li>Verify that all personnel with administrative access to the system at any point in the lifecycle meet the necessary personnel security requirements <ul><li>For example, some electoral systems may require that administrative access be given only to Canadian citizens with a valid security clearance</li> </ul></li> <li>Maintain formal security vetting for all personnel with administrative access throughout the lifecycle <ul><li>Require personnel security screening and background checks for personnel requiring privileged or administrative access to the election and voting infrastructure; software developers, hardware assembly technicians, program managers, and election administrators responsible for administering federal elections will require a security clearance to support the voting infrastructure</li> <li>Ensure administrative accounts are used exclusively to perform administrative tasks and that these privileges are regularly audited</li> </ul></li> <li>Employ two-person control for all interactions with vote tabulators during their lifecycle, except for the moment when individual voters cast ballots</li> <li>Do not share access credentials; ensure that each user has unique credentials to access vote tabulation systems</li> <li>Enforce the use of passphrases or strong passwords on voting systems</li> <li>Use strong authentication protocols and never store passwords in plain text</li> <li>Protect authentication parameters with modern hashing algorithms to provide resilience against cryptographic attacks</li> <li>Require multi-factor authentication (MFA) mechanisms for administrative access to vote tabulation systems</li> </ul><p>For more information on account and access management, read <a href="/en/guidance/best-practices-passphrases-and-passwords-itsap30032">Best practices for passphrases and passwords (ITSAP.30.032)</a>, <a href="/en/guidance/secure-your-accounts-and-devices-multi-factor-authentication-itsap30030">Secure your accounts and devices with multi-factor authentication (ITSAP.30.030)</a>, and <a href="/en/guidance/top-10-it-security-actions-no3-managing-controlling-administrative-privileges-itsm10094">Top 10 <abbr title="information technology">IT</abbr> security actions: No. 3 Managing and controlling administrative privileges (ITSM.10.094)</a>.</p> <h4>Lock down devices</h4> <p>Elections administrators should ensure that the configuration and security settings shipped with vote tabulators are updated to align with organizational policies. Elections administrators should also consider taking the following actions to lock down vote tabulators:</p> <ul><li>Restrict the services enabled on the systems and devices, and prevent users from creating or changing system settings or configurations <ul><li>Only enable services that are required for performing intended tasks</li> </ul></li> <li>Disconnect external devices, such as printers, fax systems, and computing devices, from voting tabulators when they are not in use</li> <li>Restrict access to any existing auxiliary or data ports <ul><li>If there is an operational need to enable these ports, it should be evaluated through a risk assessment review process</li> </ul></li> </ul><h3 id="5.7">Validate and authenticate devices</h3> <p>It is important to implement robust authentication mechanisms to ensure that vote tabulation devices remain trusted. Consider mechanisms to continuously validate and authenticate the integrity of vote tabulators through their lifecycle. For example, confirm the validity of trusted public key infrastructure (PKI) digital signatures on software packages. Elections administrators should consider software logic and accuracy testing to validate that the vote tabulators operate as expected. Using device cryptographic certificates can also help with physical device certification.</p> <h3 id="5.8">Patches and system updates</h3> <p>It is essential to undertake regular patch management to protect the vote tabulator against known and unknown threats. You should also implement the following measures:</p> <ul><li>Ensure device firmware and software are kept up to date</li> <li>Apply device patches and software updates regularly</li> <li>Use cryptographic mechanisms to verify firmware and software before applying updates</li> <li>Validate that update deployment cycles do not operationally impact scheduled election events</li> </ul><h3 id="5.9">Secure application development</h3> <p>Elections administrators may require software application development or procurement to meet specific legal or bylaw requirements. Security considerations should be evaluated throughout all phases of the development or procurement process. Regardless of the software application acquisition model selected, security requirements should be clearly defined and documented to assess functional and security objectives. Threat modeling or threat assessment techniques can also be used to identify and address potential threats associated with software applications to be deployed on the vote tabulator. Modern software stacks are so complex that lifecycle compromises should be assumed. Elections agencies can reduce their attack surface by identifying and managing risks early during the application design stage and architect to adequately mitigate those risks.</p> <h4>Software application testing</h4> <p>Static and dynamic testing approaches may be used to test the security of system software. Static testing involves source and binary code testing, while dynamic testing involves testing the application in functional use. Your software assurance process should allow you to evaluate applications in a test environment. Adopt functional and penetration testing controls to validate security objectives and address all identified vulnerabilities. Do not assume that an application has been vetted; instead, conduct your assessment in line with your security objectives. Leverage formal testing standards such as those defined by <a href="https://www.cyber.gc.ca/en/tools-services/common-criteria">Common Criteria</a>, <a href="https://owasp.org/www-project-application-security-verification-standard/">OWASP Application Security Verification Standard (ASVS)</a>, and <a href="https://www.iso.org/obp/ui/en/#iso:std:iso-iec:27034:-1:ed-1:v1:en">ISO/IEC 27034-1 Information technology — Security techniques ― Application security</a>.</p> <h3 id="5.10">Audit log monitoring</h3> <p>Enable event and log collection capabilities on the vote tabulator to assist with incident investigations. Activity events should be logged in protected storage, like write-once media, and retention of log data should adhere to legal and privacy regulations. Make sure to coordinate logging activities with incident response and forensic objectives to ensure that the activity logs collected contain the information needed to investigate incidents.</p> <h3 id="5.11">Business continuity</h3> <p>You should establish and test business continuity plans and procedures to ensure your election activities can recover from adverse events. Business planning should identify which data must be backed up and establish testing and recovery procedures for your back-ups. In addition, you should consider which combination of onsite or offsite back-up options best meet your needs. Back-ups require the same security protections as actual devices, including physical security, network isolation, controlled access, and two-person-control.</p> <p>For more information on business continuity planning, read <a href="/en/guidance/developing-your-business-continuity-plan-itsap10005">Developing your business continuity plan (ITSAP.10.005).</a> For more information on back-ups, read <a href="/en/guidance/tips-backing-your-information-itsap40002">Tips for backing up your information (ITSAP.40.002)</a>.</p> <p>For live elections, ensure you have contingency procedures and can quickly recover from unplanned events such as vote tabulator failures or theft. You should secure access to and encrypt sensitive data back-ups. You should also back up encryption keys, security certificates and credentials to facilitate recovery when required and disconnect external storage when not in use to minimize the likelihood that back-up data will be corrupted.</p> <h4>Retain paper backups</h4> <p>When designing elections processes that use vote tabulators, it is important that administrators consider resilient solutions to protect against disruption attacks and electronic manipulation. While keeping electronic data back-ups is a critical step in enabling quick recovery from unplanned incidents, paper records or copies of voter ballots and tabulated results should be maintained. If a catastrophic incident impacts the functioning of electoral systems, paper records will ensure the democratic process can proceed.</p> <h4>Plan for incidents and recovery</h4> <p>Security incidents may still occur even if you implement the best cyber security. Elections administrators should be prepared to deal with and recover from cyber security compromises. Have an incident recovery plan in place which outlines roles and responsibilities of all stakeholders. Test your plan regularly to assure its effectiveness. Electoral processes that rely on electronic systems should have manual and off-the-grid back-up measures that can be deployed should electronic systems fail.</p> <p>For more information on incident response and recovery planning, read <a href="/en/guidance/developing-your-incident-response-plan-itsap40003">Developing your incident response plan (ITSAP.40.003)</a> and <a href="/en/guidance/developing-your-it-recovery-plan-itsap40004">Developing your <abbr title="information technology">IT</abbr> recovery plan (ITSAP.40.002)</a>.</p> <h3 id="5.12">Educate users</h3> <p>Educating elections staff and volunteers about cyber security risks is critical for maintaining the safe operation of vote tabulators. Frontline elections staff and volunteers should be trained on how their actions could help facilitate or thwart a cyber security attack. Ensure elections staff receive regular training to understand the threat environment. Training and awareness programs should target all users of electoral systems. You should also periodically evaluate the effectiveness of your training programs and ensure the content is aligned with threat assessment feedback and addresses key threat issues.</p> <p>For more information on tailored cyber security training, read <a href="https://www.cyber.gc.ca/en/guidance/top-10-it-security-actions-6-provide-tailored-cyber-security-training-itsm10093">Top 10 <abbr title="information technology">IT</abbr> security actions: No. 6 Provide tailored cyber security training (ITSM.10.093)</a>.</p> <h4>Personnel security</h4> <p>An effective security screening process for your staff and volunteers is crucial. Implement a personnel and volunteer screening process to ascertain and validate the integrity of people recruited to work as poll staff or elections officials. Conduct security background checks before employees and volunteers are hired. Ensure the screening process continuously assesses the suitability of employees or volunteers even as they change roles or positions.</p> <p>To limit the possibility of foreign legal compulsion, only Canadian citizens with the appropriate security clearance should be granted administrative access to the vote tabulation system during any phase of its system lifecycle.</p> <h3 id="5.13">Security control effectiveness</h3> <p>Assessing and testing the effectiveness of security controls deployed on vote tabulators is an important step in ensuring systems are functioning as expected and will be resilient against real-world attacks. Elections administrators can leverage controlled testing strategies such as procedural or physical penetration testing and threat modelling techniques. In addition, the following are some other measures that can help ensure that security controls remain effective:</p> <ul><li>Automating assessment measures to continuously validate that operational, technical, and management control measures are functioning as expected</li> <li>Identifying relevant security, operational and performance metrics that should be evaluated</li> <li>Assessing the cyber resiliency capabilities of the system to withstand disruptive events such as a logic bomb</li> </ul><!–** TOP OF PAGE ******–><div class="clearfix"> </div> <div class="pull-right small text-muted mrgn-bttm-0"><a href="#wb-tphp">Top of page</a> <span aria-hidden="true" class="text-primary glyphicon glyphicon-circle-arrow-up"></span></div> <!–** END TOP OF PAGE **–> <h2 id="6">Conclusion</h2> <p>Vote tabulation devices provide electronic capabilities that can enable and enhance the democratic process and increase efficiency. However, potential threats to vote tabulators, such as disruptions or altered data, can negatively impact trust in the democratic process. The applications and deployment architecture of these systems must therefore be carefully considered. To lessen the impact of potential attacks when deploying and using vote tabulators, elections management bodies should adopt a rigorous focus on security, including appropriate control protections.</p> <p>Elections administrators must adopt a data-centric strategy which implements security controls across the data path to secure voting systems. They should ensure that vote tabulators used during democratic elections are validated and that appropriate security protections are in place. Elections administrators should further ensure that the integrity of software, firmware, and hardware deployed in tabulation devices is validated, and should continuously protect these devices and systems throughout their entire lifecycle.</p> <hr /><h2>Effective date</h2> <p>This publication takes effect on August 6, 2026.</p> <p>This is an <span class="text-uppercase">UNCLASSIFIED</span> publication that has been issued under the authority of the Head of the Canadian Centre for Cyber Security (Cyber Centre). For more information, contact the Cyber Centre:</p> <ul><li>by email: <a href="mailto:contact@cyber.gc.ca">contact@cyber.gc.ca</a></li> <li>by phone: <a href="tel:+1-613-949-7048">613-949-7048</a> or <a href="tel:+1-833-292-3788">1-833-CYBER-88</a></li> </ul></div> </div> </div> </div> </div> </article>

Discover how CSE’s Frontier AI Lab is exploring the ways artificial intelligence can strengthen cyber defence, beginning with detection engineering.

<article data-history-node-id="705" about="/en/guidance/protect-your-organization-malware-itsap00057" class="cccs-basic-page full clearfix"> <div class="content"> <div class="layout layout–onecol"> <div class="layout__region layout__region–content"> <div data-block-plugin-id="extra_field_block:node:cccs_basic_page:links" class="block block-layout-builder block-extra-field-blocknodecccs-basic-pagelinks clearfix"> </div> <div data-block-plugin-id="field_block:node:cccs_basic_page:body" class="block block-layout-builder block-field-blocknodecccs-basic-pagebody clearfix"> <div class="field field–name-body field–type-text-with-summary field–label-hidden field–item"><div class="row "> <div class="col-md-4 pull-left hidden-xs hidden-sm"> <p class="text-left"><strong>August 2026</strong></p> </div> <div class="col-md-4 hidden-xs hidden-sm"> <p class="text-center"><strong>Awareness series</strong></p> </div> <div class="col-md-4 pull-right hidden-xs hidden-sm"> <p class="text-right"><strong>ITSAP.00.057</strong></p> </div> <!–MOBILE STARTS HERE–> <div class="hidden-lg hidden-md text-center"> <p><strong>August 2026 | Awareness series</strong></p> </div> </div> <div class="clearfix"> </div> <p>Malware, also known as malicious software, is used by threat actors to compromise networks, systems and devices. When malware infects an organization, threat actors may be able to gain access to sensitive information, monitor activity and disrupt operations.</p> <p>This publication outlines common types of malware, the warning signs of an infected device and the steps to take when a device is compromised. It also provides practical tips to help protect your organization from future threats.</p> <section><h2>On this page</h2> <ul><li><a href="#1">How malware is introduced</a></li> <li><a href="#2">Malware types and behaviour</a></li> <li><a href="#3">How artificial intelligence is transforming malware</a></li> <li><a href="#4">Signs of an infected device</a></li> <li><a href="#5">Steps to address infected devices</a></li> <li><a href="#6">Tips to protect against malware</a></li> <li><a href="#7">Learn more</a></li> </ul></section><h2 id="1">How malware is introduced</h2> <p>Malware can enter a network, system or device through vulnerabilities or risky actions. The following examples highlight common ways in which users might inadvertently introduce malware into systems and devices.</p> <h3>Downloads and files</h3> <ul><li>Opening malicious email attachments</li> <li>Downloading software, apps or updates from untrusted sources</li> <li>Downloading or sharing files through peer-to-peer or unofficial file-sharing platforms</li> <li>Connecting unverified or unscanned removable media (for example, USBs or external drives)</li> </ul><h3>Browsing and access</h3> <ul><li>Clicking deceptive pop-ups, ads or fake warnings</li> <li>Installing unauthorized browser extensions or toolbars</li> <li>Visiting unsafe or compromised websites</li> <li>Using unsecured or untrusted Wi-Fi networks</li> </ul><h3>Device and account security</h3> <ul><li>Using outdated, unsupported or unpatched software or devices</li> <li>Using weak, default or reused passwords</li> </ul><h2 id="2">Malware types and behaviour</h2> <p>Malware appears in many forms, each designed to exploit systems in different ways. Understanding the different types of malware can help you detect malicious activity early and respond with confidence. Malware can be described by how it spreads and what it does once it is running. A single attack may involve more than one method and more than one capability.</p> <h3>How malware spreads</h3> <p>Malware can spread through different techniques that allow it to install itself and move between systems or networks. The types below describe how malware gains access and propagates.</p> <ul><li><strong>Trojan horse:</strong> disguises itself as harmless software to persuade users to install it, enabling unauthorized access or harmful actions</li> <li><strong>Virus:</strong> infects files and spreads to other systems through user actions, such as opening or sharing infected files</li> <li><strong>Worm:</strong> executes independently and self-replicates, usually through network connections, to cause damage such as deleting files, sending unauthorized emails or taking up bandwidth</li> </ul><h3>What malware can do</h3> <p>Once malware is running, it may perform a range of harmful actions that affect systems, data or users. The malware types below describe the impact or effects of malware on infected devices.</p> <ul><li><strong>Adware:</strong> tracks Internet activity to deliver targeted pop-up advertisements <ul><li>it is often installed without user consent through bundled software or malicious websites</li> <li>it may redirect browsing traffic or collect user data</li> </ul></li> <li><strong>Browser hijacker:</strong> manipulates browser settings to redirect users to unwanted websites or ads, often changing the homepage, modifying search defaults or adding unauthorized toolbars</li> <li><strong>Botnet:</strong> connects a network of infected devices (called "bots" or "zombies")</li> <li><strong>Cryptojacking:</strong> hijacks a device’s processing power to mine cryptocurrency without the owner’s knowledge, often causing significant slowdowns and overheating</li> <li><strong>Logic bomb:</strong> triggers malicious actions on a system when specific conditions are met, such as a particular date, time or user action, and often remains hidden until activated</li> <li><strong>Rootkit:</strong> grants hidden access to networks, systems or devices by masking itself as a legitimate operating system (OS) component. <ul><li>it frequently embeds itself deep within OS components or firmware to avoid detection</li> </ul></li> <li><strong>Ransomware:</strong> denies access to data or systems, typically by encrypting data and withholding the decryption keys until payment is made to the threat actor</li> <li><strong>Spyware:</strong> designed to monitor user activity and collect sensitive information, including browsing data, login details or personal information. Common types of spyware include: <ul><li><strong>information stealers:</strong> extract sensitive data from a device, including user credentials, browser history, browser session cookies and tokens, autofill information such as saved payment card details, communication logs, documents, system information, and even screenshots</li> <li><strong>keyloggers:</strong> capture and record keystrokes so threat actors can obtain sensitive information such as passwords, personal data or financial information <ul><li>they often operate within the OS or within keyboard input software to collect this information discreetly</li> </ul></li> </ul></li> <li><strong><abbr title="virtual private network">VPN</abbr>filter: </strong>compromises routers to allow threat actors to intercept traffic, harvest information, exploit connected systems and disrupt or manipulate network traffic</li> <li><strong>Wiper:</strong> destroys data by permanently erasing, overwriting or corrupting files and system components, often rendering devices inoperable and leaving little or no chance of recovery</li> </ul><h2 id="3">How artificial intelligence is transforming malware</h2> <p>Threat actors are taking advantage of artificial intelligence (AI) to make common types of malware more effective. The following sections outline the three ways <abbr title="artificial intelligence">AI</abbr> is transforming malware.</p> <h3>Method 1: <abbr title="artificial intelligence">AI</abbr>-assisted malware</h3> <p>This form of malware relies on <abbr title="artificial intelligence">AI</abbr> tools that operate outside the malware itself to assist in how attacks are developed or delivered. This can make malicious activity faster, more targeted and more convincing.</p> <p>For example, threat actors may use <abbr title="artificial intelligence">AI</abbr> to generate messages, craft malicious code, automated scanning of public information to identify valuable targets and craft tailored social engineering content designed to increase infection rates.</p> <p>Entry points include phishing emails with malicious links or attachments, downloads from compromised websites and the use of stolen or weak credentials to log into systems.</p> <h3>Method 2: <abbr title="artificial intelligence">AI</abbr>-augmented malware</h3> <p>This form of malware uses <abbr title="artificial intelligence">AI</abbr> to strengthen traditional malicious functions during execution to enhance stealth, speed or decision making. <abbr title="artificial intelligence">AI</abbr> enhances these actions but does not completely control or replace the underlying malware processes.</p> <p>For example, threat actors may use <abbr title="artificial intelligence">AI</abbr> to choose less visible routes for lateral movement, shape data exfiltration to resemble routine traffic, prioritize targets based on observed activity or automatically fine-tune tactics to avoid malware detection tools.</p> <p>Entry points include phishing emails with malicious links or attachments, exploitation of unpatched systems and misuse of remote access, which the malware then builds on using <abbr title="artificial intelligence">AI</abbr> to improve how it spreads or hides.</p> <h3>Method 3: Embedded <abbr title="artificial intelligence">AI</abbr> malware</h3> <p>This form of malware is designed with <abbr title="artificial intelligence">AI</abbr> embedded into the payload so it can analyze the system it infects and adjust its behaviour in real time. This allows it to identify targets, respond to security controls and choose actions that maximize its effectiveness.</p> <p>For example, <abbr title="artificial intelligence">AI</abbr> may be used to select high-value files to steal, alter activity when security tools are detected, adjust execution timing based on user behaviour or learn from system conditions to refine future actions (for example, self-modifying code).</p> <p>Entry points include phishing emails with malicious links or attachments and vulnerable or misconfigured Internet-facing services that allow threat actors to place the malware on a device or system.</p> <h2 id="4">Signs of an infected device</h2> <p>Malware often disrupts devices in subtle ways before the damage becomes obvious. Staying aware of the following indicators can help you identify potential threats early and protect your information effectively.</p> <h3>Changes on your device</h3> <ul><li>Pop-up windows appearing on your device</li> <li>Homepage changes</li> <li>Browser redirects, new or unknown toolbars or unexpected icons</li> </ul><h3>Performance issues</h3> <ul><li>Slow computer performance</li> <li>Page or system crashes</li> <li>Unusual or excessive hard drive activity, especially when the system is idle</li> </ul><h3>Security or software changes</h3> <ul><li>Unknown programs running on your device</li> <li>Antivirus or malware protection software being disabled</li> <li>Security settings or administrative controls being locked or altered</li> <li>Files becoming inaccessible, encrypted or renamed</li> <li>Ransom notes or payment demands appearing on the device</li> </ul><h3>Unusual account activity</h3> <ul><li>Spam emails or messages sent from your account</li> <li>Unauthorized password changes or unusual login activity</li> </ul><h3>Network and device behaviour</h3> <ul><li>High network traffic when the device is idle</li> <li>Overheating, rapid battery drain or unusual data use on your mobile device</li> </ul><h2 id="5">Steps to address infected devices</h2> <p>When a device becomes infected with malware, quick action is essential to limit damage, prevent further spread and protect sensitive information. The following steps outline current best practices to help you contain the threat, clean your device safely and restore it to a secure state.</p> <h3>Immediate actions</h3> <ul><li>Contact your information technology (IT) security service desk or <abbr title="information technology">IT</abbr> provider immediately</li> <li>Disconnect the infected device from all networks (for example, Wi-Fi, Ethernet, and mobile data)</li> <li>If malware appears to be active or spreading, power off the device to prevent further damage</li> <li>Refrain from using the infected device to sign into sensitive accounts, such as email, financial systems or administrative portals</li> </ul><h3>Containment and recovery</h3> <ul><li>Run antivirus scans offline when possible</li> <li>Restore data only from known, clean backups that were created before the infection</li> <li>Scan all backup files before restoring them to ensure they are free of malware <ul><li>Use trusted malware analysis tools, such as <a href="/en/tools-services/assemblyline">Assemblyline</a>, the Canadian Centre for Cyber Security’s malware detection and analysis platform, to help assess suspicious files before restoration</li> </ul></li> <li>Reset all passwords associated with the device, revoke active sign-in sessions and access tokens and re-enrol multi-factor authentication (MFA) to prevent misuse of stolen credentials</li> </ul><h3>Post-incident monitoring</h3> <ul><li>Reconnect the device to your network only after the <abbr title="operating system">OS</abbr> is reinstalled and all malware scans confirm the device is clean</li> <li>Continue monitoring the device for unusual behaviour, suspicious traffic or new alerts</li> <li>Perform regular antivirus scans to ensure no malware remains</li> </ul><h2 id="6">Tips to protect against malware</h2> <p>Staying safe online requires more than just good habits; it requires intentional and proactive cyber hygiene. Malware threats continue to evolve, targeting personal devices, corporate systems and everything in between. By following best practices such as our <a href="/en/guidance/top-10-it-security-actions">Top 10 <abbr title="information technology">IT</abbr> security actions</a> and using trusted protective tools, you can significantly reduce your risk of compromise. The following tips outline the most effective steps you can take to strengthen your defences and help keep your information, devices and accounts secure.</p> <h3>Authentication and identity security</h3> <ul><li>Enable passkeys or phishing-resistant MFA for all accounts, choosing authentication methods that align with the sensitivity of the account and the level of risk</li> <li>Use a password manager to generate and store long, unique passwords</li> <li>Limit user and administrator access to only what is required for their roles and regularly remove excess permissions</li> </ul><h3>Device and system hardening</h3> <ul><li>Install software, <abbr title="operating system">OS</abbr> and firmware updates immediately</li> <li>Use anti-virus, anti-malware and endpoint detection and response tools</li> <li>Use a firewall to block and restrict unauthorized network access</li> <li>Enable full-disk or file-based encryption</li> <li>Use host intrusion detection systems where supported</li> <li>Use application allowlisting so only authorized apps can run</li> <li>Turn off wireless features like Wi-Fi, Bluetooth, GPS and near-field communication when not in use</li> </ul><h3>Network and Internet safety</h3> <ul><li>Avoid public Wi-Fi or use a virtual private network when public Wi-Fi is necessary</li> <li>Use anti-phishing protections and align email systems with Domain-based Message Authentication, Reporting, and Conformance (DMARC)</li> <li>Use an ad blocker to reduce risk from malicious ads or a tracker blocker and limit the collection of browsing activity information</li> <li>Verify files, links and attachments before downloading or opening them</li> <li>Provide regular security awareness training so users can recognize phishing emails, malicious links, fake websites and social engineering attempts</li> <li>Encourage users to report suspected phishing or malware quickly to reduce spread and impact</li> <li>Use a protective domain name system to prevent access to malicious or suspicious domains</li> <li>Monitor device data usage for suspicious or unusual activity</li> </ul><h3>Backup and data protection</h3> <ul><li>Follow the 3-2-1 backup rule: 3 copies, 2 types of media, 1 offsite</li> <li>Encrypt sensitive data stored in backups</li> </ul><h3>Zero-trust and enterprise controls</h3> <ul><li>Implement zero-trust architecture</li> <li>Use identity threat detection and response tools</li> <li>Strengthen supply chain security and monitor third-party risks</li> </ul><h2 id="7">Learn more</h2> <ul><li><a href="/en/guidance/ransomware-playbook-itsm00099">Ransomware playbook (ITSM.00.099)</a></li> <li><a href="/en/guidance/spotting-malicious-email-messages-itsap00100">Spotting malicious email messages (ITSAP.00.100)</a></li> <li><a href="/en/guidance/how-updates-secure-your-device-itsap10096">How updates secure your device (ITSAP.10.096)</a></li> <li><a href="/en/guidance/best-practices-passphrases-and-passwords-itsap30032">Best practices for passphrases and passwords (ITSAP.30.032)</a></li> <li><a href="/en/guidance/password-managers-security-itsap30025">Password managers: Security tips (ITSAP.30.025)</a></li> <li><a href="/en/guidance/protective-domain-name-system-itsap40019">Protective Domain Name System (ITSAP.40.019)</a></li> <li><a href="/en/guidance/preventative-security-tools-itsap00058">Preventative security tools (ITSAP.00.058)</a></li> <li><a href="/en/guidance/dont-take-bait-recognize-and-avoid-phishing-attacks">Don’t take the bait: Recognize and avoid phishing attacks (ITSAP.00.101)</a></li> </ul></div> </div> </div> </div> </div> </article>