Google GTIG recorded in its AI Threat Tracker, published on Nov. 16, that threat actors are using LLM frameworks to find and exploit vulnerabilities end to end without human involvement. The exploit_agent it identified compares code before and after patches, generates exploits, and executes them automatically, primarily targeting n-days for which patches already exist. ARTEX activity disclosed contemporaneously by the FBI and Amazon shows that intrusions can gain privileged access within three hours, but these frameworks also leave traces in traffic, identifiers, and code.
Source:Techaeris · Read the original
Leave a Reply