Database/AI/ML frameworks & serving
nexent: unauthenticated document-deletion endpoint wipes ElasticSearch indices and backing MinIO objects
Impact
The ElasticSearch service interface accepts DELETE /{index_name}/documents with no authentication and no validation of the path_or_url parameter, so an unauthenticated caller chooses what to delete. Deletion propagates past the index into the MinIO objects behind it, which means the source files are gone, not just the search view. For an operator running this as the retrieval tier of an inference service, the loss is the knowledge corpus plus whatever else shares that bucket, and the service is down until it is reingested. Integrity and availability only - the record describes no read path.
Who can reach it
Anyone who can reach the nexent backend API over the network. No authentication is required.
What to do
The record names no fixed version for nexent 1.7.5.2 and points only at the project repository. Until a release lands, keep the backend API off any tenant-reachable or internet-reachable network, require authentication in front of it, and confirm you have restorable backups of the ElasticSearch indices and the MinIO bucket. Track CVE-2026-31216 at the same time - it is the same missing-auth pattern on the storage API.
References
Related entries
- nexent: unauthenticated storage API deletes arbitrary objects from the backing MinIO storeCVE-2026-31216 · nexent backend (DELETE /storage/{object_name:path} file management API)Critical
- LiteLLM (JWT auth): Auth bypass when `enable_jwt_auth` is setCVE-2026-35030 · LiteLLM (JWT auth)Critical
- vLLM: ASGI request handling lets callers bypass API-key authentication on the OpenAI endpointsCVE-2026-48746 · vLLM OpenAI-compatible API server (AuthenticationMiddleware)Critical
- SGLang (multimodal runtime): Unauthenticated path traversalCVE-2026-7302 · SGLang (multimodal runtime)Critical
- Ollama (GGUF model loader): Heap out-of-bounds read from an attacker-supplied GGUF via `/api/create`CVE-2026-7482 · Ollama (GGUF model loader)Critical
- Cocos AI: TDX aTLS verifier ignores expected REPORT_DATA, accepting replayed attestation evidenceCVE-2026-92701 · Cocos AI (Intel TDX aTLS attestation verifier, VerifTeeAttestation)Critical
This entry is curated: imported from vendor advisories with machine assistance, not yet individually verified. Confirm against your vendor's advisory before acting, and report anything wrong.