DeutschShown in its original language — a full translation is planned for a future version.
vllm: Mehrere Schwachstellen ermöglichen
view advisory published 2026-09-13
Ein Angreifer kann mehrere Schwachstellen in vllm ausnutzen, um Sicherheitsvorkehrungen zu umgehen, Daten zu manipulieren, sensible Informationen offenzulegen oder einen Denial-of-Service-Zustand auszulösen.
01
describes (8)
Related CVEs.
| ID | severity | exploitable | ||
|---|---|---|---|---|
| CVE-2026-100654US | medium6.5 | — | ||
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 a… | ||||
| CVE-2026-100653US | medium6.5 | — | ||
vLLM is an inference and serving engine for large language models. In versions from 0.22.1 through 0.28.0, the operator-supplied model revision pin (--revision / --code-revision) i… | ||||
| CVE-2026-100652US | high7.5 | — | ||
vLLM versions 0.22.0 through 0.23.0 fail to validate stop_token_ids against vocabulary bounds in Rust HTTP and gRPC frontends, allowing out-of-vocabulary token IDs to reach MinToke… | ||||
| CVE-2026-100651US | medium6.5 | — | ||
vLLM before 0.29.0 fails to enforce decoder prompt-length validation on the disaggregated serving endpoint /inference/v1/generate. When the request contains a 'features' (multimoda… | ||||
| CVE-2026-100650US | medium6.5 | — | ||
vLLM through 0.29.0 fetches and fully materializes remote or inline media before enforcing its documented media controls (the VLLM_MAX_AUDIO_CLIP_FILESIZE_MB compressed-audio size… | ||||
| CVE-2026-100649US | medium5.3 | — | ||
vLLM before 0.29.0 contains a resource-limit bypass vulnerability in PyNvVideoCodec decoder allocation where sampler subclass shadowing allows independent counter increments. Unaut… | ||||
| CVE-2026-100648US | medium5.3 | — | ||
vllm before 0.29.0 fails to enforce VLLM_MAX_AUDIO_CLIP_FILESIZE_MB limit in multimodal chat audio decoding, allowing unauthenticated clients to bypass file size restrictions. Atta… | ||||
| CVE-2026-100647US | high7.5 | — | ||
vLLM versions before 0.29.0 contain a denial-of-service vulnerability in the cache_salt parameter accepted on OpenAI-compatible and Anthropic API endpoints, which lacks maximum len… | ||||
02
references
External citations — patches, advisories, analysis.
- Red Hat Security Advisory RHSA-2026:90440 vom 2026-10-26external
- GitHub Security Advisory GHSA-3mqx-f33v-vgp9 vom 2026-09-13external
- GitHub Security Advisory GHSA-hhv2-872h-628q vom 2026-09-13external
- GitHub Security Advisory GHSA-j682-9xp5-rrf3 vom 2026-09-13external
- GitHub Security Advisory GHSA-jcq2-4gch-5qhf vom 2026-09-13external
- GitHub Security Advisory GHSA-p6g9-7v3x-m8mv vom 2026-09-13external
- GitHub Security Advisory GHSA-qff2-492f-9fm4 vom 2026-09-13external
- GitHub Security Advisory GHSA-v5gm-qgmv-gc6c vom 2026-09-13external
scry
Advisories like this name the exposed products attackers go after on your perimeter. See what is exposed on your perimeter with a free passive scan.
Free scan