Xinference through 4.0.1 contains a path traversal vulnerability on Windows that allows unauthenticated attackers to write and delete arbitrary .json files via backslash sequences in model_name. Attackers can register a persisted custom model through POST /v1/model_registrations/{model_type} with names like ..\..\evil to overwrite configuration or data files writable by the Xinference process.
Fix
Nothing published yet: no patch, vendor advisory or mitigation reference on NVD, no vendor advisory in the News for this CVE, not in CISA's KEV, no fixed package version. BaselineWatch does not guess at one.
Change history 2
- StatusReceived → DeferredNVD modified 2026-10-19 21:14
- First seenCVSS 8.3NVD modified 2026-10-19 17:17
Affected (CPE criteria) 0
NVD has not attached CPE configurations yet.
References 5
- https://github.com/xorbitsai/inference/blob/v4.0.1/xinference/model/cache_manager.py#L117-L134
- https://github.com/xorbitsai/inference/blob/v4.0.1/xinference/model/utils.py#L1175-L1182
- https://github.com/xorbitsai/inference/commit/beddc97ac211f8414597fe8479493be0603f7c84
- https://github.com/xorbitsai/inference/issues/5702
- https://www.vulncheck.com/advisories/xinference-through-4.0.1-path-traversal-via-model-name-in-custom-model-registration