Описание
vLLM is an inference and serving engine for large language models (LLMs). Prior to 0.22.0, vLLM's revision pinning controls do not consistently apply to all artifacts loaded for a model. A deployment that supplies --revision or --code-revision can still load dynamic code, GGUF files, image processors, retrieval side weights, or same-repository subfolder weights/config from an unpinned/default revision. This is a supply-chain integrity issue for pinned vLLM deployments. Operators can believe they are serving a reviewed model revision while vLLM resolves behavior-affecting nested or sibling artifacts outside that reviewed revision. This vulnerability is fixed in 0.22.0.
A flaw was found in vLLM, an inference and serving engine for large language models (LLMs). The revision pinning controls in vLLM do not consistently apply to all artifacts loaded for a model. This allows a deployment configured with specific revisions to still load dynamic code or other configuration files from an unpinned or default revision. This issue can lead to a supply-chain integrity compromise, where operators may unknowingly serve models with unreviewed or unintended behavior.
Отчет
Red Hat rates this issue as having Moderate impact. The flaw is a supply-chain integrity issue when operators pin a HuggingFace model revision but vLLM may still load nested artifacts from an unpinned revision. It affects Red Hat AI Inference Server, Red Hat OpenShift AI, and Red Hat Enterprise Linux AI images that ship vLLM versions prior to 0.22.0. KServe control-plane components that bundle vLLM as a library are not affected.
Меры по смягчению последствий
Upgrade to a vLLM build containing the fix (>= 0.22.0) when available from Red Hat. Until then, only serve models from trusted registries, pin revisions explicitly, and review nested artifacts in model repositories before deployment.
Затронутые пакеты
| Платформа | Пакет | Состояние | Рекомендация | Релиз |
|---|---|---|---|---|
| Red Hat AI Inference Server | rhaiis/vllm-cpu-rhel9 | Fix deferred | ||
| Red Hat AI Inference Server | rhaiis/vllm-cuda-rhel9 | Fix deferred | ||
| Red Hat AI Inference Server | rhaiis/vllm-neuron-rhel9 | Fix deferred | ||
| Red Hat AI Inference Server | rhaiis/vllm-rocm-rhel9 | Fix deferred | ||
| Red Hat AI Inference Server | rhaiis/vllm-spyre-rhel9 | Fix deferred | ||
| Red Hat AI Inference Server | rhaiis/vllm-tpu-rhel9 | Fix deferred | ||
| Red Hat AI Inference Server | rhaii/vllm-cpu-rhel9 | Fix deferred | ||
| Red Hat AI Inference Server | rhaii/vllm-cuda-rhel9 | Fix deferred | ||
| Red Hat AI Inference Server | rhaii/vllm-gaudi-rhel9 | Fix deferred | ||
| Red Hat AI Inference Server | rhaii/vllm-neuron-rhel9 | Fix deferred |
Показывать по
Ссылки на источники
Дополнительная информация
Статус:
EPSS
6.5 Medium
CVSS3
Связанные уязвимости
vLLM is an inference and serving engine for large language models (LLMs). Prior to 0.22.0, vLLM's revision pinning controls do not consistently apply to all artifacts loaded for a model. A deployment that supplies --revision or --code-revision can still load dynamic code, GGUF files, image processors, retrieval side weights, or same-repository subfolder weights/config from an unpinned/default revision. This is a supply-chain integrity issue for pinned vLLM deployments. Operators can believe they are serving a reviewed model revision while vLLM resolves behavior-affecting nested or sibling artifacts outside that reviewed revision. This vulnerability is fixed in 0.22.0.
vLLM is an inference and serving engine for large language models (LLM ...
vLLM's Artifact Pin Decay allows pinned deployments to load unpinned code, weights, and processors
EPSS
6.5 Medium
CVSS3