Aller au contenu
en Se connecter
7.1 Élevée

CVE-2026-100654 vLLM before 0.29.0 accepts user-controlled stop_token_ids on the OpenAI-compatible POST /v1/completions and POST…

Description

vLLM before 0.29.0 accepts user-controlled stop_token_ids on the OpenAI-compatible POST /v1/completions and POST /v1/chat/completions endpoints but validates only that the values are integers, not that each token id is within the model vocabulary/logits range. When min_tokens > 0, the stop token ids are used as logits indices to suppress stop tokens, so an out-of-range id reaches a CUDA indexing operation (index_put_) and triggers a device-side assertion. An authenticated API user can send a single malformed completion request that returns 500 Internal Server Error and puts EngineCore into a fatal state, causing subsequent requests to fail until the service is restarted (denial of service).

Produits et versions touchés

Produit Versions touchées Corrigé en
vllm 0.29 premières → < 0.29.0 0.29.0

Références

Chronologie

  1. Publiée
  2. Dernière modification

Autres CVE de vllm

Les 121 CVE de vllm →
CVE-2026-115918 vLLM is an inference and serving engine for large language models 6.5 Moyenne CVE-2026-115917 vLLM is an inference and serving engine for large language models 3.4 Faible CVE-2026-115916 vLLM is an inference and serving engine for large language models 6.5 Moyenne CVE-2026-115915 vLLM is an inference and serving engine for large language models 8.1 Élevée
Suivre cette CVE