Unit42 Palo Alto Networks

We discuss vulnerabilities in popular GenAI web products to LLM jailbreaks. Single-turn strategies remain effective, but multi-turn approaches show greater success. The post Investigating LLM Jailbreaking of Popular Generative AI Web Products appeared first on Unit 42.

Unit 42 details the just-discovered connection between threat group Stately Taurus (aka Mustang Panda) and the malware Bookworm, found during analysis of the group’s infrastructure. The post Stately Taurus Activity in Southeast Asia Links to Bookworm Malware appeared first on Unit 42.

Unit 42 researchers detail nine vulnerabilities discovered in NVIDIA’s CUDA-based toolkit. The affected utilities help analyze cubin (binary) files. The post Multiple Vulnerabilities Discovered in NVIDIA CUDA Toolkit appeared first on Unit 42.

Atomic Stealer, Poseidon Stealer and Cthulhu Stealer target macOS. We discuss their various properties and examine leverage of the AppleScript framework. The post Stealers on the Rise: A Closer Look at a Growing macOS Threat appeared first on Unit 42.

Evaluation of three jailbreaking techniques on DeepSeek shows risks of generating prohibited content. The post Recent Jailbreaks Demonstrate Emerging Threat to DeepSeek appeared first on Unit 42.

A Chinese-linked espionage campaign targeted entities in South Asia using rare techniques like DNS exfiltration, with the aim to steal sensitive data. The post CL-STA-0048: An Espionage Operation Against High-Value Targets in South Asia appeared first on Unit 42.

CVE-2025-0282 and CVE-2025-0283 affect multiple Ivanti products. This threat brief covers attack scope, including details from an incident response case. The post Threat Brief: CVE-2025-0282 and CVE-2025-0283 (Updated Jan. 17) appeared first on Unit 42.

Graph neural networks aid in analyzing domains linked to known attack indicators, effectively uncovering new malicious domains and cybercrime campaigns. The post One Step Ahead in Cyber Hide-and-Seek: Automating Malicious Infrastructure Discovery With Graph Neural Networks appeared first on Unit 42.

The jailbreak technique “Bad Likert Judge” manipulates LLMs to generate harmful content using Likert scales, exposing safety gaps in LLM guardrails. The post Bad Likert Judge: A Novel Multi-Turn Technique to Jailbreak LLMs by Misusing Their Evaluation Capability appeared first on Unit 42.

This article demonstrates how AI can be used to modify and help detect JavaScript malware. We boosted our detection rates 10% with retraining. The post Now You See Me, Now You Don’t: Using LLMs to Obfuscate Malicious JavaScript appeared first on Unit 42.

A phishing campaign targeting European companies used fake forms made with HubSpot’s Free Form Builder, leading to credential harvesting and Azure account takeover. The post Effective Phishing Campaign Targeting European Companies and Organizations appeared first on Unit 42.

Using real-world examples and offering plenty of pragmatic tips, learn how to protect your directory services from LDAP-based attacks. The post LDAP Enumeration: Unveiling the Double-Edged Sword of Active Directory appeared first on Unit 42.

Vulnerabilities in Microsoft Azure Data Factory’s integration with Apache Airflow can lead to unauthorized access and control over cloud resources. The post Dirty DAG: New Vulnerabilities in Azure Data Factory’s Apache Airflow Integration appeared first on Unit 42.

Analysis of packer-as-a-service (PaaS) HeartCrypt reveals its use in over 2k malicious payloads across 45 malware families since its early 2024 appearance. The post Crypted Hearts: Exposing the HeartCrypt Packer-as-a-Service Operation appeared first on Unit 42.

Unit 42 probes network abuses around events like the Olympics, featuring case studies of scams and phishing through domain registrations and more. The post Network Abuses Leveraging High-Profile Events: Suspicious Domain Registrations and Other Scams appeared first on Unit 42.