arxiv.org Natural Code Rewrites Exploit Gaps in CodeQL Security Analysis
Article Content
- •Natural code rewrites can silence CodeQL alerts while leaving vulnerabilities intact.
- •26.2% of tested vulnerability flows were silenced by maintenance refactorings.
- •Seven specific gaps in library modeling were identified as contributing to this issue.
On November 12, 2026, researchers from North Carolina State University published findings revealing that natural code rewrites can suppress security alerts from GitHub's CodeQL static analysis tool while leaving vulnerabilities exploitable. Their study evaluated 172 real-world vulnerability flows in popular npm packages, finding that 26.2% of cases were silenced by maintenance refactorings generated by a large language model. This indicates a significant flaw in how static application security testing (SAST) tools interpret alerts, as developers often assume that disappearing alerts mean vulnerabilities are resolved. The researchers identified seven specific gaps in how static analyzers model runtime libraries, with the majority of successful rewrites exploiting these gaps. The findings raise concerns about the reliability of automated security gates in continuous integration workflows.
Ask AI about this cluster
Answers cite the sources they use