Description
vLLM before 0.29.0 validates allowed_token_ids against tokenizer length instead of model output logits width in SamplingParams._validate_allowed_token_ids(). Attackers can supply token IDs above the output vocabulary that pass validation, causing LogitBiasState to corrupt GPU logits state and allow concurrent requests to sample tokens outside their allowlists.
Metrics
CVSS 3.1 5.3/10
CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:L/A:NWeakness enumeration
Affected software
No affected versions have been published.
Timeline
- Published
- Sep 18, 2026
- Last modified
- Sep 28, 2026
- NVD status
- Analyzed