Bug 2558029 (CVE-2026-114049) - CVE-2026-114049 vllm: vLLM: Output manipulation via stale token filter reuse in batch processing
Summary: CVE-2026-114049 vllm: vLLM: Output manipulation via stale token filter reuse ...
Keywords:
Status: NEW
Alias: CVE-2026-114049
Product: Security Response
Classification: Other
Component: vulnerability
Version: unspecified
Hardware: All
OS: Linux
low
low
Target Milestone: ---
Assignee: Product Security DevOps Team
QA Contact:
URL:
Whiteboard:
Depends On:
Blocks:
TreeView+ depends on / blocked
 
Reported: 2026-10-21 01:01 UTC by OSIDB Bzimport
Modified: 2026-10-22 11:26 UTC (History)
5 users (show)

Fixed In Version:
Clone Of:
Environment:
Last Closed:
Embargoed:

Attachments (Terms of Use)

Description OSIDB Bzimport 2026-10-21 01:01:53 UTC 
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.
Note You need to log in before you can comment on or make changes to this bug.