{"id":"ghsa-ghsa-h3rc-6mm3-gc2m","seq":1708793,"kind":"advisory","title":"vLLM: Request-controlled mm_processor_kwargs.code_revision allows remote code execution","url":"https://github.com/advisories/GHSA-h3rc-6mm3-gc2m","permalink":"https://vulnpulse.app/items/ghsa-ghsa-h3rc-6mm3-gc2m","summary":"影響: pip:vllm >= 0.7.3, < 0.31.0 → 0.31.0\nAn untrusted multimodal request can set mm_processor_kwargs.code_revision. When vLLM serves a model whose Hugging Face processor uses dynamic auto_map code and the operator enabled --trust-remote-code, this request value reaches AutoProcessor.from_pretrained and selects the Python revision imported by Transformers. This lets a client replace the operator-reviewed processor module with code from another revision of the same model repository. The operator-selected model/data revision remains pinn…","severity":"high","techs":["python"],"cves":["CVE-2026-114053"],"tags":["GHSA-h3rc-6mm3-gc2m","vllm"],"source":{"id":"ghsa","name":"GitHub Advisory Database"},"published_at":"2026-10-27T15:36:17.000Z","fetched_at":"2026-10-27T21:03:52.214Z"}