Database/AI/ML frameworks & serving
vLLM (multimodal embeddings): Memory corruption
Impact
Memory corruption → crash and possible RCE
Who can reach it
Unauthenticated request to an exposed OpenAI-compatible serving port carrying crafted embeddings
What to do
Upgrade to 0.11.1+. If the provider hosts the endpoint, this is provider-owned; if the tenant runs their own vLLM, it is tenant-owned and the provider can only enforce network policy
Fleet impact
How widespread
Very common - vLLM is the default open-source inference server on nearly every neocloud's serverless-inference and BYO-endpoint product
Cost to remediate
daemon-restart - upgrade to 0.11.1+ and roll the serving fleet; every replica of every model endpoint must be restarted
Why it hits the whole fleet
The Completions API torch.loads user-supplied prompt embeddings; with PyTorch 2.8's sparse-tensor checks off by default a crafted tensor causes an out-of-bounds write - network-only, unauthenticated-adjacent, on every serving replica
References
Related entries
- Fugue RPC server: unauthenticated cloudpickle deserialization gives code execution on cluster workersCVE-2025-62703 · Fugue FlaskRPCServer (cloudpickle deserialization in fugue/rpc/flask.py)High
- Jupyter Server: gateway kernel channel skips the execute authorization check, letting restricted users run codeCVE-2026-114135 · Jupyter Server (gateway WebSocketChannelsHandler kernel authorization)High
- nbconvert: notebook output metadata filename escapes output_files_dir, writing files anywhere the converter can reachCVE-2026-114140 · Jupyter nbconvert (ExtractOutputPreprocessor output filename handling)High
- vLLM: forged shared-memory handles in multimodal kwargs reach pickle.loads on workersCVE-2026-115912 · vLLM (shared-memory multimodal feature cache)High
- vLLM: attacker-supplied X-Request-Id makes a prefill worker ZMQ-connect out, leaking NCCL id and KV cacheCVE-2026-116176 · vLLM P2pNcclConnector (decode address from X-Request-Id header)High
- OpenShift AI dashboard: incorrect network binding lets an in-cluster actor impersonate any user with an arbitrary tokenCVE-2026-16745 · Red Hat OpenShift AI odh-dashboard (network binding / token handling)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.