Database/AI/ML frameworks & serving
NVIDIA OpenShell: HTTP request smuggling past the L7 REST proxy's policy checks
Impact
OpenShell's L7 REST proxy and the upstream server disagree on where a request body ends. A sandboxed workload can use that to hide a second request inside the first, and the hidden request skips the proxy's policy checks. That allows tampering with upstream services the sandbox should not modify (integrity high, scope changed). Attack complexity is high.
Who can reach it
Low-privileged actor inside an OpenShell sandbox, such as agent code. No user interaction needed.
What to do
Update NVIDIA OpenShell to the fixed release in NVIDIA product-security bulletin 5929. The record does not give a fixed version. Until then, enforce authorization on the upstream services, rather than relying on the proxy alone.
References
Related entries
- Jupyter Server: login `next` parameter allows redirect to an arbitrary external hostCVE-2025-61669 · Jupyter Server (LoginFormHandler redirect validation)Medium
- vLLM: video decoder limit bypass via sampler subclass shadowing exhausts unaccounted GPU memoryCVE-2026-100649 · vLLM (PyNvVideoCodec decoder allocation, sampler subclass accounting)Medium
- docling-serve: memory telemetry endpoints omit the auth dependency, exposing stats and forced GCCVE-2026-108749 · docling-serve (/v1/memory/stats and /v1/memory/counts endpoints)Medium
- vLLM: stale allowed_token_ids rows let one request constrain another tenant's outputCVE-2026-114049 · vLLM InputBatch.condense() (allowed_token_ids cross-request mask leak)Medium
- MLflow: path traversal in model_id lets unauthenticated clients write files outside the run directoryCVE-2026-127202 · MLflow tracking server (FileStore backend, log outputs API)Medium
- TrustyAI Service Operator: unauthenticated access to AI guardrail and orchestrator APIsCVE-2026-15044 · TrustyAI Service Operator (Red Hat OpenShift AI)Medium
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.