Reproduce “Guiding Symbolic Execution with Static Analysis and LLMs for Vulnerability Discovery” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Md Shafiuzzaman, Achintya Desai, Wenbo Guo, Tevfik Bultan

The Sailor pipeline automates the discovery of memory-safety vulnerabilities in large C/C++ codebases by integrating static analysis, large language models, and symbolic execution with concrete validation. It confirmed 379 unique, previously unknown vulnerabilities across 7 open-source projects.

Key insights

  • Automating symbolic execution harness generation via LLMs and an iterative feedback loop is crucial for applying symbolic execution at scale, overcoming its traditional scalability bottlenecks.
  • A synergistic combination of static analysis (for scalable targeting), LLMs (for code synthesis), symbolic execution (for precision), and concrete validation (for ground truth) is necessary to achieve both scale and accuracy in vulnerability discovery.
  • Raw LLM-based vulnerability detection or static analysis alone yield very high false positive rates (88-99%), underscoring the necessity of precise execution and concrete validation to filter findings and confirm feasibility.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.