{"api_version":"1","generated_at":"2026-07-04T18:19:03+00:00","cve":"CVE-2026-14535","urls":{"html":"https://cve.report/CVE-2026-14535","api":"https://cve.report/api/cve/CVE-2026-14535.json","docs":"https://cve.report/api","cve_org":"https://www.cve.org/CVERecord?id=CVE-2026-14535","nvd":"https://nvd.nist.gov/vuln/detail/CVE-2026-14535"},"summary":{"title":"Fickling MLAllowlist analysis pass rendered inoperative by shared mutable state in AnalysisContext.shorten_code()","description":"In Trail of Bits fickling versions up to and including 0.1.11, the UnsafeImportsML analysis pass unconditionally calls AnalysisContext.shorten_code(node) on every import node it inspects, regardless of whether the import is flagged as unsafe. This call registers the shortened code representation in the shared AnalysisContext.reported_shortened_code set. When the MLAllowlist analysis pass subsequently runs, it calls the same shorten_code() method, receives already_reported=True for every import, and executes a continue statement that skips its allowlist check entirely. This renders MLAllowlist dead code for all imports — it never evaluates whether an import is in the ML allowlist or not. The MLAllowlist pass was designed to catch imports of modules outside the known-safe ML ecosystem (torch, numpy, transformers, etc.) that slip past the UnsafeImports denylist. With MLAllowlist inoperative, any standard library module not in the UNSAFE_IMPORTS denylist can be invoked via pickle deserialization while fickling's check_safety() returns LIKELY_SAFE. The fickling.load() API chains check_safety() into pickle.loads() as an explicit security gate, meaning a LIKELY_SAFE verdict causes the payload to be deserialized and executed. The root cause is shared mutable state between independently-correct analysis passes — UnsafeImportsML works as designed in isolation, MLAllowlist works as designed in isolation, but the shared reported_shortened_code set causes UnsafeImportsML to poison MLAllowlist's deduplication logic.","state":"PUBLISHED","assigner":"BombadilSystems","published_at":"2026-07-04 14:16:29","updated_at":"2026-07-04 14:16:29"},"problem_types":["CWE-693","CWE-693 CWE-693 Protection Mechanism Failure"],"metrics":[{"version":"3.1","source":"aa17e1a1-c329-4d6e-a1ed-8d0188aea082","type":"Secondary","score":"8.8","severity":"HIGH","vector":"CVSS:3.1/AV:N/AC:L/PR:N/UI:R/S:U/C:H/I:H/A:H","data":{"version":"3.1","vectorString":"CVSS:3.1/AV:N/AC:L/PR:N/UI:R/S:U/C:H/I:H/A:H","baseScore":8.8,"baseSeverity":"HIGH","attackVector":"NETWORK","attackComplexity":"LOW","privilegesRequired":"NONE","userInteraction":"REQUIRED","scope":"UNCHANGED","confidentialityImpact":"HIGH","integrityImpact":"HIGH","availabilityImpact":"HIGH"}},{"version":"3.1","source":"CNA","type":"CVSS","score":"8.8","severity":"HIGH","vector":"CVSS:3.1/AV:N/AC:L/PR:N/UI:R/S:U/C:H/I:H/A:H","data":{"attackComplexity":"LOW","attackVector":"NETWORK","availabilityImpact":"HIGH","baseScore":8.8,"baseSeverity":"HIGH","confidentialityImpact":"HIGH","integrityImpact":"HIGH","privilegesRequired":"NONE","scope":"UNCHANGED","userInteraction":"REQUIRED","vectorString":"CVSS:3.1/AV:N/AC:L/PR:N/UI:R/S:U/C:H/I:H/A:H","version":"3.1"}}],"references":[{"url":"https://github.com/trailofbits/fickling/commit/41ce7cb01edd97072994039574a2301ebb3f463d","name":"https://github.com/trailofbits/fickling/commit/41ce7cb01edd97072994039574a2301ebb3f463d","refsource":"aa17e1a1-c329-4d6e-a1ed-8d0188aea082","tags":[],"title":"","mime":"","httpstatus":"","archivestatus":"0"},{"url":"https://github.com/trailofbits/fickling/security/advisories/GHSA-cffv-grgg-g429","name":"https://github.com/trailofbits/fickling/security/advisories/GHSA-cffv-grgg-g429","refsource":"aa17e1a1-c329-4d6e-a1ed-8d0188aea082","tags":[],"title":"","mime":"","httpstatus":"","archivestatus":"0"},{"url":"https://github.com/trailofbits/fickling/pull/278","name":"https://github.com/trailofbits/fickling/pull/278","refsource":"aa17e1a1-c329-4d6e-a1ed-8d0188aea082","tags":[],"title":"","mime":"","httpstatus":"","archivestatus":"0"},{"url":"https://github.com/trailofbits/fickling/releases/tag/v0.1.12","name":"https://github.com/trailofbits/fickling/releases/tag/v0.1.12","refsource":"aa17e1a1-c329-4d6e-a1ed-8d0188aea082","tags":[],"title":"","mime":"","httpstatus":"","archivestatus":"0"},{"url":"https://www.cve.org/CVERecord?id=CVE-2026-14535","name":"CVE Program record","refsource":"CVE.ORG","tags":["canonical"]},{"url":"https://nvd.nist.gov/vuln/detail/CVE-2026-14535","name":"NVD vulnerability detail","refsource":"NVD","tags":["canonical","analysis"]}],"affected":[{"source":"CNA","vendor":"trailofbits","product":"fickling","version":"affected 0.1.11 custom","platforms":[]},{"source":"CNA","vendor":"trailofbits","product":"fickling","version":"unaffected 0.1.12 custom","platforms":[]}],"timeline":[],"solutions":[],"workarounds":[],"exploits":[],"credits":[{"source":"CNA","value":"Christopher Aziz (Bombadil Systems LLC)","lang":"en"}],"nvd_cpes":[],"vendor_comments":[],"enrichments":{"kev":null,"epss":null,"legacy_qids":[]},"source_records":{"cve_program":{"containers":{"cna":{"affected":[{"collectionURL":"https://pypi.org/project/fickling/","defaultStatus":"unaffected","packageName":"fickling","product":"fickling","vendor":"trailofbits","versions":[{"lessThanOrEqual":"0.1.11","status":"affected","version":"0","versionType":"custom"},{"status":"unaffected","version":"0.1.12","versionType":"custom"}]}],"credits":[{"lang":"en","type":"finder","value":"Christopher Aziz (Bombadil Systems LLC)"}],"datePublic":"2026-06-28T00:08:00.000Z","descriptions":[{"lang":"en","supportingMedia":[{"base64":false,"type":"text/html","value":"<p>In Trail of Bits fickling versions up to and including 0.1.11, the UnsafeImportsML analysis pass unconditionally calls AnalysisContext.shorten_code(node) on every import node it inspects, regardless of whether the import is flagged as unsafe. This call registers the shortened code representation in the shared AnalysisContext.reported_shortened_code set. When the MLAllowlist analysis pass subsequently runs, it calls the same shorten_code() method, receives already_reported=True for every import, and executes a continue statement that skips its allowlist check entirely. This renders MLAllowlist dead code for all imports — it never evaluates whether an import is in the ML allowlist or not. The MLAllowlist pass was designed to catch imports of modules outside the known-safe ML ecosystem (torch, numpy, transformers, etc.) that slip past the UnsafeImports denylist. With MLAllowlist inoperative, any standard library module not in the UNSAFE_IMPORTS denylist can be invoked via pickle deserialization while fickling's check_safety() returns LIKELY_SAFE. The fickling.load() API chains check_safety() into pickle.loads() as an explicit security gate, meaning a LIKELY_SAFE verdict causes the payload to be deserialized and executed. The root cause is shared mutable state between independently-correct analysis passes — UnsafeImportsML works as designed in isolation, MLAllowlist works as designed in isolation, but the shared reported_shortened_code set causes UnsafeImportsML to poison MLAllowlist's deduplication logic.</p>"}],"value":"In Trail of Bits fickling versions up to and including 0.1.11, the UnsafeImportsML analysis pass unconditionally calls AnalysisContext.shorten_code(node) on every import node it inspects, regardless of whether the import is flagged as unsafe. This call registers the shortened code representation in the shared AnalysisContext.reported_shortened_code set. When the MLAllowlist analysis pass subsequently runs, it calls the same shorten_code() method, receives already_reported=True for every import, and executes a continue statement that skips its allowlist check entirely. This renders MLAllowlist dead code for all imports — it never evaluates whether an import is in the ML allowlist or not. The MLAllowlist pass was designed to catch imports of modules outside the known-safe ML ecosystem (torch, numpy, transformers, etc.) that slip past the UnsafeImports denylist. With MLAllowlist inoperative, any standard library module not in the UNSAFE_IMPORTS denylist can be invoked via pickle deserialization while fickling's check_safety() returns LIKELY_SAFE. The fickling.load() API chains check_safety() into pickle.loads() as an explicit security gate, meaning a LIKELY_SAFE verdict causes the payload to be deserialized and executed. The root cause is shared mutable state between independently-correct analysis passes — UnsafeImportsML works as designed in isolation, MLAllowlist works as designed in isolation, but the shared reported_shortened_code set causes UnsafeImportsML to poison MLAllowlist's deduplication logic."}],"impacts":[{"descriptions":[{"lang":"en","value":"An attacker can craft a malicious pickle file that imports any standard library module not in fickling's UNSAFE_IMPORTS denylist. Because the MLAllowlist check is silently skipped due to shared state pollution from UnsafeImportsML, fickling's check_safety() returns LIKELY_SAFE. When used with fickling.load(), the payload is deserialized and arbitrary code executes in the victim's environment. This is distinct from denylist incompleteness (CWE-184) because the allowlist mechanism was specifically designed to catch imports that the denylist misses, and a code defect renders it entirely inoperative."}]}],"metrics":[{"cvssV3_1":{"attackComplexity":"LOW","attackVector":"NETWORK","availabilityImpact":"HIGH","baseScore":8.8,"baseSeverity":"HIGH","confidentialityImpact":"HIGH","integrityImpact":"HIGH","privilegesRequired":"NONE","scope":"UNCHANGED","userInteraction":"REQUIRED","vectorString":"CVSS:3.1/AV:N/AC:L/PR:N/UI:R/S:U/C:H/I:H/A:H","version":"3.1"},"format":"CVSS","scenarios":[{"lang":"en","value":"GENERAL"}]}],"problemTypes":[{"descriptions":[{"cweId":"CWE-693","description":"CWE-693 Protection Mechanism Failure","lang":"en","type":"CWE"}]}],"providerMetadata":{"dateUpdated":"2026-07-04T13:31:14.937Z","orgId":"aa17e1a1-c329-4d6e-a1ed-8d0188aea082","shortName":"BombadilSystems"},"references":[{"name":"GitHub Security Advisory GHSA-cffv-grgg-g429","url":"https://github.com/trailofbits/fickling/security/advisories/GHSA-cffv-grgg-g429"},{"name":"Fix PR #278 — Split shorten_code() formatting from dedup tracking, make MLAllowlist opt-in","url":"https://github.com/trailofbits/fickling/pull/278"},{"name":"Fix commit 41ce7cb (squash merge of PR #278)","url":"https://github.com/trailofbits/fickling/commit/41ce7cb01edd97072994039574a2301ebb3f463d"},{"name":"Fickling v0.1.12 release","url":"https://github.com/trailofbits/fickling/releases/tag/v0.1.12"}],"source":{"discovery":"EXTERNAL"},"title":"Fickling MLAllowlist analysis pass rendered inoperative by shared mutable state in AnalysisContext.shorten_code()","x_generator":{"engine":"Vulnogram 1.0.2"}}},"cveMetadata":{"assignerOrgId":"aa17e1a1-c329-4d6e-a1ed-8d0188aea082","assignerShortName":"BombadilSystems","cveId":"CVE-2026-14535","datePublished":"2026-07-04T13:31:14.937Z","dateReserved":"2026-07-03T00:02:55.918Z","dateUpdated":"2026-07-04T13:31:14.937Z","state":"PUBLISHED"},"dataType":"CVE_RECORD","dataVersion":"5.2"},"nvd":{"publishedDate":"2026-07-04 14:16:29","lastModifiedDate":"2026-07-04 14:16:29","problem_types":["CWE-693","CWE-693 CWE-693 Protection Mechanism Failure"],"metrics":{"cvssMetricV31":[{"source":"aa17e1a1-c329-4d6e-a1ed-8d0188aea082","type":"Secondary","cvssData":{"version":"3.1","vectorString":"CVSS:3.1/AV:N/AC:L/PR:N/UI:R/S:U/C:H/I:H/A:H","baseScore":8.8,"baseSeverity":"HIGH","attackVector":"NETWORK","attackComplexity":"LOW","privilegesRequired":"NONE","userInteraction":"REQUIRED","scope":"UNCHANGED","confidentialityImpact":"HIGH","integrityImpact":"HIGH","availabilityImpact":"HIGH"},"exploitabilityScore":2.8,"impactScore":5.9}]},"configurations":[]},"legacy_mitre":{"record":{"CveYear":"2026","CveId":"14535","Ordinal":"1","Title":"Fickling MLAllowlist analysis pass rendered inoperative by share","CVE":"CVE-2026-14535","Year":"2026"},"notes":[{"CveYear":"2026","CveId":"14535","Ordinal":"1","NoteData":"In Trail of Bits fickling versions up to and including 0.1.11, the UnsafeImportsML analysis pass unconditionally calls AnalysisContext.shorten_code(node) on every import node it inspects, regardless of whether the import is flagged as unsafe. This call registers the shortened code representation in the shared AnalysisContext.reported_shortened_code set. When the MLAllowlist analysis pass subsequently runs, it calls the same shorten_code() method, receives already_reported=True for every import, and executes a continue statement that skips its allowlist check entirely. This renders MLAllowlist dead code for all imports — it never evaluates whether an import is in the ML allowlist or not. The MLAllowlist pass was designed to catch imports of modules outside the known-safe ML ecosystem (torch, numpy, transformers, etc.) that slip past the UnsafeImports denylist. With MLAllowlist inoperative, any standard library module not in the UNSAFE_IMPORTS denylist can be invoked via pickle deserialization while fickling's check_safety() returns LIKELY_SAFE. The fickling.load() API chains check_safety() into pickle.loads() as an explicit security gate, meaning a LIKELY_SAFE verdict causes the payload to be deserialized and executed. The root cause is shared mutable state between independently-correct analysis passes — UnsafeImportsML works as designed in isolation, MLAllowlist works as designed in isolation, but the shared reported_shortened_code set causes UnsafeImportsML to poison MLAllowlist's deduplication logic.","Type":"Description","Title":"Fickling MLAllowlist analysis pass rendered inoperative by share"}]}}}