CVE-2026-66007

MEDIUM

Datasets Path Traversal via Unsanitized file_name Metadata

Title source: cna
STIX 2.1

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.

References (4)

Core 4
Core References
Exploit technical-description exploit issue-tracking
Researcher Disclosure
https://github.com/huggingface/datasets/issues/8324
Issue Tracking issue-tracking patch
Pull Request
https://github.com/huggingface/datasets/pull/8325

Scores

CVSS v3 6.5
EPSS 0.0052
EPSS Percentile 41.0%
Attack Vector NETWORK
CVSS:3.1/AV:N/AC:L/PR:N/UI:R/S:U/C:H/I:N/A:N

CISA SSVC

Vulnrichment
Exploitation poc
Automatable no
Technical Impact partial

Details

CWE
CWE-22
Status published
Products (2)
huggingface/datasets < 5.0.0
huggingface/datasets f989ef9b4cc6c0039a7a82458eebca49e2b58b4b
Published Jul 24, 2026
Tracked Since Jul 24, 2026