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.
2026年10月21日
CVE-2026-114049 [LOW] – vLLM before 0.31.0 contains a cross-request mask leak in InputBatch.condense() that allows
📑 文章目录



