MediumVulnerability
GHSA-25rm-9wvm-m38v: falcon-rust: Discrete-Gaussian sampler runs at 2^-33 precision, far below Falcon's requirement (reference KAT vectors no longer pass)
- Identifiers
- CVE-2026-77382GHSA-25rm-9wvm-m38v
- Published
- Record updated
- Affected
- falcon-rust <= 0.1.3
- Fixed in
- 0.2.0
- released July 12, 2026, 131 days before disclosure
- Known exploitation
- Not listed in the CISA Known Exploited Vulnerabilities catalog at the last check.
Summary
falcon-rust v0.1.3 and the current master (HEAD cfa1b35) evaluate the discrete-Gaussian sampler's exp() acceptance step at only 32 fractional bits, in src/samplerz.rs (approx_exp, ber_exp, sampler_z), introduced by commit bf2cf00. Measured maximum relative error is 2^-33, far coarser than the reference implementation's roughly 2^-63 and the replaced f64 code's roughly 2^-52, and below the 2^-40 to 2^-46 range the Falcon security analysis is cited as requiring. The crate also fails Falcon's known-answer test vectors, which the maintainers addressed by loosening the test tolerance.
Mitigation
The source does not state a fix yet. Check the original advisory for updates.
Related items
- InfoRECLAIM: Robot-Vision-Informed Adversarial Patch Defense via Embodied Attention FeedbackSimilar attack · Arxiv (cs.RO + cs.CV security)
- LowExecGap: Exploiting Command-Execution Gap for Patch Attacks on Vision-Language-Action ModelsSimilar attack · Arxiv (cs.RO + cs.CV security)
- InfoConfGuard: a physics-grounded reference monitor for predictive-maintenance large-language-model agentsSimilar attack · Arxiv (cs.CR + cs.CL + cs.LG)
- MediumGemini AI agents repeatedly bypassed safeguards to attack real websiteSimilar attack · BleepingComputer
- InfoWatermark Evidence under Open-Weight LLM Modification: Exact Bounds for Low-Rank Output WatermarksSimilar attack · Arxiv (cs.CR + cs.CL + cs.LG)