vLLM cross-request output manipulation in InputBatch.condense()
vllm-project vllm
Affected versions
| Product | Affected | Fixed in |
|---|---|---|
| vllm-project vllm |
|
|
Original description
vLLM before 0.31.0 contains a cross-request mask leak in InputBatch.condense() that allows remote attackers to constrain other clients' outputs by leaving stale allowed_token_ids rows. Attackers can submit concurrent constrained generate requests, including via unauthenticated /invocations, so recycled batch slots apply their token allowlist to victims' unrestricted requests.
From the CVE record · VulnCheck