CVE-2026-93989

Publié : 19 septembre 2026

Modifié : 19 septembre 2026

Score CVSS

3.1

LOW

Description détaillée

vLLM through 0.29.0 fails to properly validate bad_words token indices against the model's generation output width in SamplingParams.update_from_tokenizer(). Attackers can supply out-of-bounds token indices that corrupt logits memory of concurrent requests, causing different in-flight HTTP requests to return incorrect tokens.

Vecteur d'attaque (CVSS)

Vecteur brut :CVSS:3.1/AV:N/AC:H/PR:L/UI:N/S:U/C:N/I:L/A:N

Références et Patchs

https://github.com/vllm-project/vllm https://github.com/vllm-project/vllm/blob/98dff2a81d747d1dba01a47f939f48c3526d4206/vllm/sampling_params.py#L694-L753 https://github.com/vllm-project/vllm/blob/98dff2a81d747d1dba01a47f939f48c3526d4206/vllm/v1/worker/gpu/sample/bad_words.py https://github.com/vllm-project/vllm/pull/48824 https://www.vulncheck.com/advisories/vllm-through-0.29.0-cross-request-logits-corruption-via-bad-words

Dernières Vulnérabilités

CVE-2026-94057

Exim before 4.100.1 allows SMTP smuggling in which the received message does not match any sent message, and instead depends on crafted data sent after a rejection during DATA processing.

VOIR DÉTAILS

CVE-2026-94056

Exim before 4.100.1, when Proxy-Protocol is used with an attacker-controlled proxy, allows attackers to read certain uninitialized data from stack memory.

VOIR DÉTAILS

CVE-2026-94055

Exim before 4.100.1, when certain non-default TLS settings are used with GnuTLS, has a use-after-free.

VOIR DÉTAILS