Post

Can AI invent new attack techniques? New research from James Kettle and PortSwigger Research

HTTP Terminator is an autonomous AI-driven attack discovery system that invents new HTTP-based attack techniques by parsing 138 technical specifications into 15,000 inspirational fragments and generating 30,000 unique attack vectors for real-time testing. The system confirmed approximately 700 vulnerable targets across government infrastructure, financial institutions, and enterprise products through authorized bug bounty testing. The research demonstrates that while AI excels at generating attack candidates and executing repetitive testing at scale, human expertise remains critical at the discovery cascade phase—where findings become hypotheses for deeper investigation. The system successfully advanced HTTP desync attack research beyond four years of prior work, showing that autonomous methodology can amplify researcher effectiveness rather than replace human judgment. James Kettle has released both the full technical paper and HTTP Terminator source code as an open-source proof-of-concept and blueprint for other researchers.

Read original article

This post is licensed under CC BY 4.0 by the author.