AUTO-UPDATED

http-terminator – AI-Assisted HTTP Request-Smuggling Discovery

PortSwigger has released http-terminator, an open-source research pipeline that utilizes large language models to analyze HTTP specifications and automatically generate test cases for identifying request-smuggling vulnerabilities.

Key Points

  • The pipeline consists of four distinct stages: seeker, flamer, validator, and investigator.
  • seeker and flamer use Anthropic’s Claude model to extract desync vectors from documentation and convert them into malformed HTTP test cases.
  • The validator stage functions as a Burp Suite extension to test targets, while the investigator stage uses Claude Code to replicate and document findings.
  • The repository serves as a reference design for document-driven security testing rather than a standalone, turn-key product.
  • Full implementation requires external dependencies, including Burp Suite, specific Java and Python environments, and an Anthropic API key.

Why it Matters

This project demonstrates a novel, specification-first approach to automating the discovery of complex HTTP desync vulnerabilities. By providing an inspectable framework, PortSwigger offers security researchers a blueprint for integrating generative AI into specialized vulnerability research pipelines.
Darknet.org.uk Published by Darknet
Read original