aiml-code-vuln-benchmark

What problem does it solve?

Builds a robust framework to generate and validate realistic vulnerability samples across six OWASP classes, enabling automated evaluation of AI-generated exploit code and PoC payloads.

Core Features & Use Cases

• Generates vulnerability samples for sql_injection, xss_stored, rce_command_injection, path_traversal, ssrf, and insecure_deserialization.
• Enforces concrete vulnerable code and proof-of-concept exploits, minimum code length, and vulnerability sinks to ensure actionable benchmarks.
• Useful for testing static analyzers, red-teaming tooling, and model evaluation pipelines that require anchored vulnerability data.

Quick Start

Generate a complete vulnerability sample by replacing all placeholders ('???') with real vulnerable code and a working PoC exploit.