Signal is active with 1 mentions in latest observed window
Recommended action window: Monitor and triage in normal cycle
NVD description
vLLM from 0.8.5 before 0.31.0 has a denial of service flaw. LogprobsTensors.empty_cpu() in vllm/v1/outputs.py allocates an uninitialized int32 tensor, and the V1 model runner leaves prompt positions reported as computed by a KV connector unwritten. A client sharing the engine can send prompt_logprobs (or echo with logprobs) against a connector-cached prefix, making tokenizer decode raise OverflowError, which crashes EngineCore and drops all co-tenant requests.
Priority
LOW
Exploitation
NONE
PoC
NONE
Patch
NONE
Momentum
NONE
If you run products in this scope, you should treat this CVE as relevant to your environment.
vLLM EngineCore GPU Cluster Crash (CVE-2026-116180) • Logprob index underflow during prefill • Device-side CUDA assertion failure • Forces SIGABRT crash across GPU nodes Read full Blog: https://sh3llc0d3.com/blog/vllm-enginecore-denial-of-service-and-logprob-memory-corruption-weaponizing-malformed-prompt-states-to-crash-distributed-gpu-clusters-cve-2026-116180/ #sh3llc0d3 #AISecurity #vLLM #ZeroDay
1 of 1 entries
| Part | Vendor | Product | Version | Target SW | Target HW |
|---|---|---|---|---|---|
| App | vllm | vllm | - | - | - |