Researchers also say defensive tactics can be built with large language models.
Malicious actors could leverage large language models to generate 10,000 new JavaScript codes based on existing malware samples.
According to researchers from Palo Alto Networks’ Unit 42, such newly generated malware variants would be able to circumvent detection in VirusTotal, and evade deep learning models in 88 percent of tests.
According to The Hacker News, researchers said the scale of new malicious code variants could increase with the help of generative AI. “We can use the same tactics to rewrite malicious code to help generate training data that can improve the robustness of ML models," said the researchers.
Such findings follow a report by Morphisec detailing a proof-of-concept manipulation attack aimed at the Exploit Prediction Scoring System AI framework that could derail vulnerability management initiatives.
Written by
Dan Raywood
Senior Editor
SC Media UK
Dan Raywood is a B2B journalist with more than 20 years of experience, including covering cybersecurity for the past 16 years. He has extensively covered topics from Advanced Persistent Threats and nation-state hackers to major data breaches and regulatory changes.
He has spoken at events including 44CON, Infosecurity Europe, RANT Conference, BSides Scotland, Steelcon and ESET Security Days.
Outside work, Dan enjoys supporting Tottenham Hotspur, managing mischievous cats, and sampling craft beers.
