Evidence · DiggingBeagle record
Vulnerability-preserving code rewrites evade CodeQL taint tracking in a controlled study (research paper)
An LLM guided by known static-analysis alerts made apparently ordinary JavaScript rewrites that preserved unsafe flows while removing some CodeQL warnings.
- Published
- Nov 12, 2026
- Source role
- primary disclosure
Evidence record
An LLM guided by known static-analysis alerts made apparently ordinary JavaScript rewrites that preserved unsafe flows while removing some CodeQL warnings.
Claim-level citations (4)
- supportsVulnerability-preserving code rewrites evade CodeQL taint tracking in a controlled study: The researchers attribute 43 of 70 successful vulnerability-preserving scanner-evasion rewrites to a WHATWG URL parser modeling gap and report that 64 of those 70 rewrites pass through APIs that CodeQL models at least in part. This is a specific source-to-sink coverage problem, not evidence that 43 deployed applications were compromised.
Section 6.2, evasion mechanism analysis and WHATWG URL parser/source-to-sink modeling counts
- supportsVulnerability-preserving code rewrites evade CodeQL taint tracking in a controlled study: Within three tries, the researchers silenced a targeted alert for 26.2 percent of the initially detected flows; alert-unaware benign rewrites silenced none of 516 control attempts.
Experiments, rewrite-success and control comparison results
- supportsVulnerability-preserving code rewrites evade CodeQL taint tracking in a controlled study: A model receives a vulnerable code fragment and the associated warning, then makes bounded maintenance-style edits while researchers check whether the original runtime flow remains and CodeQL stops reporting that specific path.
Approach and taint-model failure taxonomy
- supportsVulnerability-preserving code rewrites evade CodeQL taint tracking in a controlled study: Of the 70 verified alert-silencing rewrites, seven still produced another finding under the same CodeQL rule; suppressing the targeted source-to-sink flow was therefore not equivalent to clearing every finding in the function.
Section 6, RQ1 and manual verification of silenced targeted flows