exploitgym

ExploitGym is a comprehensive benchmarking platform designed to evaluate the capabilities of AI agents in exploiting a variety of real-world vulnerabilities present in userspace programs, Google's V8 engine, and the Linux kernel. It features a large-scale set of 869 instances and includes detailed setup and evaluation processes, as well as the ability to disable system defenses for more effective testing. This tool serves as an essential resource for researchers aiming to understand AI's effectiveness in security exploit development.