Datasets Path Traversal via Unsanitized file_name Metadata
huggingface datasets
Affected versions
| Product | Affected | Fixed in |
|---|---|---|
| datasets PyPI |
|
|
Original description
Datasets through 5.0.0, fixed in commit f989ef9, contains a path traversal vulnerability in folder-based dataset builders where the file_name metadata field is not properly validated before being joined to the dataset directory. Attackers can supply crafted file_name values with directory traversal sequences to read arbitrary local files, which are then embedded into output when save_to_disk or push_to_hub is called.
From the CVE record · VulnCheck
References
- github.com/huggingface/datasets/issues/8324Vendor advisory
- github.com/huggingface/datasets/pull/8325Patch
- github.com/huggingface/datasets/commit/f989ef9b4cc6c0039a7a82458eebca49e2b58b4bPatch
- vulncheck.com/advisories/datasets-path-traversal-via-unsanitized-file-name-metadataPatch
- github.com/CVEProject/cvelistV5/tree/main/cves/2026/66xxx/CVE-2026-66007.json
- nvd.nist.gov/vuln/detail/CVE-2026-66007
- github.com/huggingface/datasets
- github.com/pypa/advisory-database/tree/main/vulns/datasets/PYSEC-2026-3716.yaml
- github.com/advisories/GHSA-379c-qx7v-6h59