{"vuln_id":"CVE-2026-77382","aliases":["GHSA-25rm-9wvm-m38v"],"description":"falcon-rust: Discrete-Gaussian sampler runs at 2^-33 precision, far below Falcon's requirement (reference KAT vectors no longer pass)","published_at":"2026-11-19T17:19:00.740000Z","modified_at":"2026-11-20T00:30:06.809056Z","first_seen_at":"2026-11-21T19:15:19.309822Z","severity":"medium","severity_source":"nvd","cvss_score":5.9,"cvss_vector":"CVSS:3.1/AV:N/AC:H/PR:N/UI:N/S:U/C:H/I:N/A:N","cwe":["CWE-1240"],"vendor":null,"product":null,"kev":{"is_kev":false,"date_added":null,"due_date":null,"ransomware":false},"epss":0.00274,"epss_percentile":0.20021,"nvd_status":"Received","enriched_at":null,"links":{"nvd":"https://nvd.nist.gov/vuln/detail/CVE-2026-77382"},"matched_alias":null,"items":[{"id":11016,"source":{"id":"osv","name":"OSV.dev"},"title":"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)","summary":"falcon-rust: Discrete-Gaussian sampler runs at 2^-33 precision, far below Falcon's requirement (reference KAT vectors no longer pass)\n\n**Reporter:** Connor Webber\n**Package:** `falcon-rust` (crates.io), repository `github.com/aszepieniec/falcon-rust`\n**Affected:** v0.1.3 and the current `master` (HEAD `cfa1b35`). Regression introduced in commit `bf2cf00`.\n**Component:** discrete-Gaussian sampler, `src/samplerz.rs` (`approx_exp`, `ber_exp`, `sampler_z`)\n**Status:** Confirmed by source review and by empirical measurement of the crate's own compiled code.\n\n## Summary\n\nfalcon-rust's discrete-Gaussian rejection sampler evaluates its `exp()` acceptance step at only 32 bits of fractional precision. Measured directly against the crate's own compiled code, `approx_exp` has a maximum relative error of 2^-33. For comparison, the Falcon reference implementation achieves about 2^-63, and the f64 code this crate replaced achieved about 2^-52. The current sampler is therefore roughly 2^19 (about 500,000x) less precise than the version it replaced, and about 2^30 less precise than the reference.\n\nFalcon's security proof relies on the sampled signatures being statistically indistinguishable from an ideal discrete Gaussian, so the signature transcript leaks nothing about the secret trapdoor. That guarantee holds only when the sampler's relative error stays small enough, over the roughly 2^64 signature budget, to keep the Renyi divergence between the real and ideal distributions negligible. The precision the security analysis relies on is cited in the literature between roughly 2^-40 and 2^-46 (for example the FACCT and isochronous-sampling analyses). At 2^-33 the sampler is below every value in that range, roughly 128x to 4096x too coarse depending on which bound you take, so the precondition the proof depends on is not met.\n\nThere is a second, bound-independent piece of evidence that removes any ambiguity: the crate's sampler can no longer reproduce Falcon's known-answer test (KAT) vectors, which are the authoritative correctness oracle. Rather than restore the precision, the maintainers loosened their own test tolerance and noted that \"rejection decisions may differ at the boundary.\" The regression is already visible in-tree; its security significance was simply not recognized.\n\nBecause the sampler's sigma is derived from the secret basis Gram-Schmidt norms, the resulting distribution deviation is secret-dependent. The `exp` function itself is public, but its inputs, and therefore the deviation, depend on secret quantities. This is the class of flaw that statistical and hidden-number-problem learning attacks on biased Gaussian samplers target.\n\nThe crate's own documentation describes this sampler as \"side-channel-hardened\" and notes that \"the sampler defines the secret-key distribution, so getting it wrong weakens the cryptography,\" which is exactly the property this defect touches.\n\n## Technical detail\n\nRoot location: `src/samplerz.rs`\n\n* `FixedPoint64 = FixedPoint\n` with `FRAC_UNIT = 1 << 32`, i.e. 32 fractional bits (`src/fixed_point.rs`). The crate already defines `FixedPoint128` with 6\n\nAffected: crates.io:falcon-rust\n\nFixed in: 0.2.0\n\nReferences:\nhttps://github.com/aszepieniec/falcon-rust/security/advisories/GHSA-25rm-9wvm-m38v\nhttps://nvd.nist.gov/vuln/detail/CVE-2026-77382\nhttps://github.com/aszepieniec/falcon-rust/commit/930d766256f9d67ec9b6d437557286cf47f042ca\nhttps://github.com/aszepieniec/falcon-rust/commit/bf2cf000e63a9803f7ab1a34f629a5a0e9c49cf6\nhttps://github.com/aszepieniec/falcon-rust\nhttps://github.com/aszepieniec/falcon-rust/releases/tag/v0.2.0","url":"https://osv.dev/vulnerability/GHSA-25rm-9wvm-m38v","published_at":"2026-11-20T00:17:04Z","updated_at":"2026-11-20T00:30:06.809056Z","fetched_at":"2026-11-21T19:15:55.946371Z","severity":"medium","cvss_score":5.9,"epss":0.00274,"is_kev":false,"has_exploit":false,"kev_due_date":null,"vendor":null,"product":"falcon-rust","vuln_id":"CVE-2026-77382","vuln_ids":["CVE-2026-77382","GHSA-25RM-9WVM-M38V"],"tags":["osv","crates.io","fix-available"],"cluster_id":2074,"links":{"self":"https://vulnwatch.ve1ve2.com/api/v1/items/11016","html":"https://vulnwatch.ve1ve2.com/item/11016"}},{"id":2325,"source":{"id":"ghsa","name":"GitHub Security Advisories"},"title":"GHSA-25rm-9wvm-m38v / CVE-2026-77382: falcon-rust: Discrete-Gaussian sampler runs at 2^-33 precision, far below Falcon's requirement (reference KAT vectors no longer pass)","summary":"falcon-rust: Discrete-Gaussian sampler runs at 2^-33 precision, far below Falcon's requirement (reference KAT vectors no longer pass)\n\n**Reporter:** Connor Webber\n**Package:** `falcon-rust` (crates.io), repository `github.com/aszepieniec/falcon-rust`\n**Affected:** v0.1.3 and the current `master` (HEAD `cfa1b35`). Regression introduced in commit `bf2cf00`.\n**Component:** discrete-Gaussian sampler, `src/samplerz.rs` (`approx_exp`, `ber_exp`, `sampler_z`)\n**Status:** Confirmed by source review and by empirical measurement of the crate's own compiled code.\n\n## Summary\n\nfalcon-rust's discrete-Gaussian rejection sampler evaluates its `exp()` acceptance step at only 32 bits of fractional precision. Measured directly against the crate's own compiled code, `approx_exp` has a maximum relative error of 2^-33. For comparison, the Falcon reference implementation achieves about 2^-63, and the f64 code this crate replaced achieved about 2^-52. The current sampler is therefore roughly 2^19 (about 500,000x) less precise than the version it replaced, and about 2^30 less precise than the reference.\n\nFalcon's security proof relies on the sampled signatures being statistically indistinguishable from an ideal discrete Gaussian, so the signature transcript leaks nothing about the secret trapdoor. That guarantee holds only when the sampler's relative error stays small enough, over the roughly 2^64 signature budget, to keep the Renyi divergence between the real and ideal distributions negligible. The precision the security analysis relies on is cited in the literature between roughly 2^-40 and 2^-46 (for example the FACCT and isochronous-sampling analyses). At 2^-33 the sampler is below every value in that range, roughly 128x to 4096x too coarse depending on which bound you take, so the precondition the proof depends on is not met.\n\nThere is a second, bound-independent piece of evidence that removes any ambiguity: the crate's sampler can no longer reproduce Falcon's known-answer test (KAT) vectors, which are the authoritative correctness oracle. Rather than restore the precision, the maintainers loosened their own test tolerance and noted that \"rejection decisions may differ at the boundary.\" The regression is already visible in-tree; its security significance was simply not recognized.\n\nBecause the sampler's sigma is derived from the secret basis Gram-Schmidt norms, the resulting distribution deviation is secret-dependent. The `exp` function itself is public, but its inputs, and therefore the deviation, depend on secret quantities. This is the class of flaw that statistical and hidden-number-problem learning attacks on biased Gaussian samplers target.\n\nThe crate's own documentation describes this sampler as \"side-channel-hardened\" and notes that \"the sampler defines the secret-key distribution, so getting it wrong weakens the cryptography,\" which is exactly the property this defect touches.\n\n## Technical detail\n\nRoot location: `src/samplerz.rs`\n\n* `FixedPoint64 = FixedPoint\n` with `FRAC_UNIT = 1 << 32`, i.e. 32 fractional bits (`src/fixed_point.rs`). The crate already defines `FixedPoint128` with 6\n\nAffected packages: rust:falcon-rust\nPatched: falcon-rust 0.2.0","url":"https://github.com/advisories/GHSA-25rm-9wvm-m38v","published_at":"2026-11-20T00:17:04Z","updated_at":"2026-11-20T00:17:04Z","fetched_at":"2026-11-21T19:15:18.416751Z","severity":"medium","cvss_score":5.9,"epss":0.00274,"is_kev":false,"has_exploit":false,"kev_due_date":null,"vendor":null,"product":"falcon-rust","vuln_id":"CVE-2026-77382","vuln_ids":["CVE-2026-77382","GHSA-25RM-9WVM-M38V"],"tags":["ghsa","rust","CWE-1240"],"cluster_id":2074,"links":{"self":"https://vulnwatch.ve1ve2.com/api/v1/items/2325","html":"https://vulnwatch.ve1ve2.com/item/2325"}},{"id":9020,"source":{"id":"nvd","name":"NVD CVE API 2.0"},"title":"CVE-2026-77382: falcon-rust is a Rust implementation of the Falcon post-quantum signature scheme. Prior to 0.2.0, the discrete-Gaussian sampler in src/samplerz.rs evaluates ap…","summary":"falcon-rust is a Rust implementation of the Falcon post-quantum signature scheme. Prior to 0.2.0, the discrete-Gaussian sampler in src/samplerz.rs evaluates approx_exp, ber_exp, and sampler_z acceptance probabilities with FixedPoint64 values having only 32 fractional bits. The scaling in approx_exp quantizes its input and limits relative precision to about 2^-33, below Falcon's required bound, causing reference known-answer vectors to diverge and invalidating the proof assumption that secret-dependent signature and key distributions are negligibly close to the ideal discrete Gaussian. This is a cryptographic-soundness defect and no practical key-recovery attack has been demonstrated. This issue is fixed in version 0.2.0.\n\nReferences:\nhttps://github.com/aszepieniec/falcon-rust/commit/930d766256f9d67ec9b6d437557286cf47f042ca\nhttps://github.com/aszepieniec/falcon-rust/commit/bf2cf000e63a9803f7ab1a34f629a5a0e9c49cf6\nhttps://github.com/aszepieniec/falcon-rust/releases/tag/v0.2.0\nhttps://github.com/aszepieniec/falcon-rust/security/advisories/GHSA-25rm-9wvm-m38v\nhttps://github.com/aszepieniec/falcon-rust/security/advisories/GHSA-25rm-9wvm-m38v","url":"https://nvd.nist.gov/vuln/detail/CVE-2026-77382","published_at":"2026-11-19T17:19:00.740000Z","updated_at":"2026-11-19T19:17:33.490000Z","fetched_at":"2026-11-21T19:15:31.484676Z","severity":"medium","cvss_score":5.9,"epss":0.00274,"is_kev":false,"has_exploit":false,"kev_due_date":null,"vendor":null,"product":null,"vuln_id":"CVE-2026-77382","vuln_ids":["CVE-2026-77382","GHSA-25RM-9WVM-M38V"],"tags":["CWE-1240"],"cluster_id":2075,"links":{"self":"https://vulnwatch.ve1ve2.com/api/v1/items/9020","html":"https://vulnwatch.ve1ve2.com/item/9020"}}]}