Database/AI/ML frameworks & serving
NVIDIA Triton Inference Server: unauthenticated request triggers excessive iteration and hangs the server
Impact
An unauthenticated request can drive Triton into excessive iteration, ending in denial of service for the inference endpoint. On a GPU fleet this takes out the process that owns the GPUs for that model, so every tenant or caller behind that endpoint loses service until the daemon is restarted, and the reserved GPUs sit idle in the meantime. NVIDIA's record scores confidentiality and integrity as unaffected, so the exposure is availability only. The advisory text does not say which request path or which parser is involved.
Who can reach it
Anyone who can reach the Triton HTTP/gRPC endpoint over the network. No authentication required (CVSS PR:N), so the blast radius depends entirely on whether the endpoint is exposed beyond the cluster or fronted by a gateway that authenticates.
What to do
Update Triton to the fixed release named in NVIDIA bulletin 2026/5875 and restart the inference server; the record given here does not state the fixed version, so read the bulletin before scheduling. No node drain or reboot is implied - this is a userspace daemon restart per serving pod, which for a replicated deployment is a rolling restart. Until patched, keep the endpoint behind an authenticating gateway and rate-limit untrusted callers.
References
Related entries
- Keras (HDF5 external links): Arbitrary local file read during model loadCVE-2026-1669 · Keras (HDF5 external links)High
- OpenShift AI Data Science Pipelines Operator: predictable PRNG generates MariaDB and MinIO credentialsCVE-2026-18611 · Red Hat OpenShift AI Data Science Pipelines Operator (generated MariaDB and MinIO credentials)High
- MLflow: unauthenticated arbitrary file read via prompt-tagged model version source pathCVE-2026-2614 · MLflow tracking server (`_create_model_version` / model version artifact handler)High
- Ray Dashboard: Path traversal in the dashboard static-file handler (port 8265)CVE-2026-32981 · Ray DashboardHigh
- LangChain: prompt config files are loaded from unvalidated paths, letting a caller read arbitrary host filesCVE-2026-34070 · LangChain langchain-core prompt loading (load_prompt / load_prompt_from_config)High
- vLLM: unbounded media download from user-supplied URLs exhausts inference server memoryCVE-2026-37237 · vLLM multimodal input fetcher (`AsyncMediaIO.fetch_audio` / `fetch_image`)High
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.