MediumVulnerability
GHSA-h668-6x6g-f8r5: tract: Arbitrary file read via unsanitized ONNX external_data `location` (path traversal) on model load in tract-onnx
- Identifiers
- CVE-2026-55832GHSA-h668-6x6g-f8r5
- Published
- Record updated
- Affected
- tract-onnx >= 0.23.0, < 0.23.2, fixed in 0.23.2, released June 16, 2026, 3 days before disclosure
- tract-onnx >= 0.22.0, < 0.22.3, fixed in 0.22.3, released June 16, 2026, 3 days before disclosure
- tract-onnx < 0.21.17, fixed in 0.21.17, released June 16, 2026, 3 days before disclosure
- Known exploitation
- Not listed in the CISA Known Exploited Vulnerabilities catalog at the last check.
- EPSS
- 0.2%
Summary
tract-onnx resolves an ONNX tensor's external-data location by joining it onto the model directory with no sanitization, so a malicious model can make tract read an arbitrary local file at load time and surface its contents in the inference output. Confirmed against tract-onnx 0.21.16 with a proof-of-concept using an absolute path, and out-of-range offset/length values can cause a panic (denial of service).
Mitigation
The source does not state a fix yet. Check the original advisory for updates.
Related items
- MediumCVE-2026-121936: easyVoice server-side request forgery through openaiBaseUrl in /api/v1/ttsSimilar attack · NVD/CVE Database
- MediumCVE-2026-118628: MLflow get-artifacts server-side request forgery information disclosureSimilar attack · NVD/CVE Database
- HighCVE-2026-118625: MLflow Tracking Server privilege escalation via get_endpointsSimilar attack · NVD/CVE Database
- InfoWhen trust becomes the payload: Inside the threats, etc quarterly reportSimilar attack · Elastic Security Labs
- MediumCritical Langflow flaws enable authentication bypass and RCESimilar attack · CSO Online