iris

IRIS is a neurosymbolic framework that integrates large language models (LLMs) with static analysis for the detection of security vulnerabilities in software projects. It analyzes codebases by taking a project and a specified Common Weakness Enumeration (CWE) type as input, and produces potential vulnerability reports along with data on their related CVEs. Notable features include the CWE-Bench-Java dataset, which contains 213 CVEs across 49 CWEs, manual extraction of source and sink specifications, and Docker integration for streamlined deployment.