AI Can Generate 10,000 Malware Variants, Study Finds
/ 1 min read
🎭 AI-Driven Malware Generation Poses New Cybersecurity Threats. Researchers from Palo Alto Networks’ Unit 42 have revealed that large language models (LLMs) can be exploited to create numerous variants of malicious JavaScript code, enhancing their ability to evade detection. While LLMs struggle to generate malware from scratch, they can effectively rewrite existing code, making it appear benign to malware classification systems. This technique can produce up to 10,000 new JavaScript variants while maintaining the original functionality, with an 88% success rate in deceiving classifiers. Additionally, a separate study from North Carolina State University demonstrated a method to extract model configurations from Google’s TPUs, highlighting the potential for intellectual property theft and further cyber attacks.
