vLLM: Scale-out disaggregated multimodal transport trusts caller-supplied features

Why it is Yellow

• CVSS 6.5 medium
• On your watchlist: PyPI

Affected packages

References

• https://github.com/advisories/GHSA-ph72-cqr5-qpp7
• https://github.com/vllm-project/vllm/security/advisories/GHSA-ph72-cqr5-qpp7
• https://github.com/vllm-project/vllm/pull/51898
• https://github.com/vllm-project/vllm/commit/1970f3ed4be7fa8620e4ddc4a12c36a8384cfc27
• https://github.com/vllm-project/vllm/releases/tag/v0.30.0
• https://github.com/advisories/GHSA-ph72-cqr5-qpp7

Original description

vLLM is an inference and serving engine for large language models. Prior to 0.30.0, the /inference/v1/generate endpoint in the disaggregated scale-out path accepts caller-supplied tensors in the features.kwargs_data field, cache identifiers in the features.mm_hashes field, ranges in the features.mm_placeholders field, and wire-selected multimodal field processors without rebinding them to the active model renderer contract. Forged grid geometry, field types, or non-positive placeholder lengths can terminate the shared EngineCore; when an attacker knows or can induce a victim's content hash, forged cache hashes can poison or retrieve cross-request encoder-cache state; and dropped sparse placeholder masks can alter replayed transport semantics. This issue is fixed in version 0.30.0.