GHSA-H3RC-6MM3-GC2M
vLLM: Request-controlled mm_processor_kwargs.code_revision allows remote code execution
Ciclo de vida de la vulnerabilidad
| Fecha | Fase | Evento | Detalle |
|---|---|---|---|
| Remediación | Parche 0.31.0 | vllm · -22 d desde divulgación | |
| Divulgación | Publicada | GHSA-H3RC-6MM3-GC2M divulgada | |
| Divulgación | Modificada | Última actualización del registro | |
| Monitoreo | Detectada | Ingresó al índice |
Summary
An 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 pinned; only the executed processor code changes.
Importing a Python module is arbitrary code execution in the vLLM API/renderer process. The PoC below demonstrates this with an import-time filesystem write. It is fully offline and does not require a GPU, model weights, or changes to a real Hub repository.
Affected snapshot
- Repository: `vllm-project/vllm`
https://github.com/vllm-project/vllm - Verified commit: `570735520942b6428ae98f1eb442c6d88adae55e`
https://github.com/vllm-project/vllm/commit/570735520942b6428ae98f1eb442c6d88adae55e - Verified with Transformers 5.15.0 and 5.16.1
- Affected release range has not been bisected
Affected configuration
The vulnerable path exists when all of the following are true:
- The served model's processor uses dynamic `auto_map.AutoProcessor` code.
- The operator enabled `trust_remote_code`.
- The server can fetch the request-selected revision, or it is already cached.
- The attacker can create or otherwise control a revision in the served model repository. An example is a Hugging Face community pull-request ref
https://huggingface.co/docs/hub/en/repositories-pull-requests-discussions#where-in-the-git-repo-are-changes-storedwhen that repository permits them. - A multimodal request reaches the processor path.
The default `trust_remote_code=False` configuration is not affected.
Attack surface
`mm_processor_kwargs` is accepted directly by public request models:
- `TokenizeChatRequest.mm_processor_kwargs`
https://github.com/vllm-project/vllm/blob/570735520942b6428ae98f1eb442c6d88adae55e/vllm/entrypoints/serve/tokenize/protocol.py#L111-L114 - `ChatCompletionRequest.mm_processor_kwargs`
https://github.com/vllm-project/vllm/blob/570735520942b6428ae98f1eb442c6d88adae55e/vllm/entrypoints/openai/chat_completion/protocol.py#L372-L375
`/tokenize` is particularly relevant because vLLM's authentication middleware guards only `/v1`, `/v2`, `/inference`, and `/cohere` prefixes:
- `AuthenticationMiddleware`
https://github.com/vllm-project/vllm/blob/570735520942b6428ae98f1eb442c6d88adae55e/vllm/entrypoints/serve/middleware/authenticate.py#L11-L62
Consequently, `/tokenize` remains unauthenticated even when `--api-key` is configured, unless an external proxy protects it. The vLLM Security Guide https://docs.vllm.ai/en/latest/usage/security/#unprotected-endpoints-no-api-key-required documents `/tokenize` as an unprotected endpoint.
Root cause
The request value follows this path:
- `OnlineRenderer.preprocess_chat`
https://github.com/vllm-project/vllm/blob/570735520942b6428ae98f1eb442c6d88adae55e/vllm/renderers/online_renderer.py#L406-L435places request `mm_processor_kwargs` in prompt extras. - Multimodal rendering forwards those kwargs into processor construction through `MMProcessorInputs`
https://github.com/vllm-project/vllm/blob/570735520942b6428ae98f1eb442c6d88adae55e/vllm/renderers/base.py#L751-L760and `_process_multimodal_async`https://github.com/vllm-project/vllm/blob/570735520942b6428ae98f1eb442c6d88adae55e/vllm/renderers/base.py#L831-L847. - `MultiModalConfig.merge_mm_processor_kwargs`
https://github.com/vllm-project/vllm/blob/570735520942b6428ae98f1eb442c6d88adae55e/vllm/config/multimodal.py#L506-L518overlays inference values on startup values, so the request wins. - `cached_get_processor_without_dynamic_kwargs`
https://github.com/vllm-project/vllm/blob/570735520942b6428ae98f1eb442c6d88adae55e/vllm/transformers_utils/processor.py#L344-L375filters only recognized processor-call keys. `code_revision` is retained as a processor-construction kwarg. - `cached_processor_from_config`
https://github.com/vllm-project/vllm/blob/570735520942b6428ae98f1eb442c6d88adae55e/vllm/transformers_utils/processor.py#L380-L390pins `revision` and `trust_remote_code`, but forwards the remaining kwargs without reserving `code_revision`. - `get_processor`
https://github.com/vllm-project/vllm/blob/570735520942b6428ae98f1eb442c6d88adae55e/vllm/transformers_utils/processor.py#L205-L245calls `AutoProcessor.from_pretrained(..., revision=revision, trust_remote_code=trust_remote_code, **kwargs)`. - Transformers `AutoProcessor.from_pretrained`
https://github.com/huggingface/transformers/blob/v5.15.0/src/transformers/models/auto/processing_auto.py#L294-L302uses `code_revision` to retrieve and import the dynamic module through `get_class_from_dynamic_module`https://github.com/huggingface/transformers/blob/v5.15.0/src/transformers/dynamic_module_utils.py#L474-L575, then removes it before loading processor data. This is why code comes from the attacker revision while configuration/data remain operator-pinned.
Request-time processor arguments are not separated from processor-construction arguments, and `code_revision` is not reserved.
Suggested remediation
- Separate processor-construction arguments from untrusted processor-call arguments. An allowlist of supported request-time preprocessing controls is safest.
- Reject `code_revision` and other loader/source controls in request `mm_processor_kwargs`, and pass `model_config.code_revision` explicitly as the only permitted processor-code revision.
- Add a regression test proving a request cannot change the loaded module revision while operator pins remain effective.
- Protect `/tokenize` with the configured authentication policy as defense in depth.
Expected safe behavior is either a 4xx response for reserved request keys or silently ignoring them before processor construction. Rejection is preferable because it makes attempted misuse visible.
A patch for this was merged here: https://github.com/vllm-project/vllm/pull/58830
Incluye detalles técnicos publicados por el investigador. Úselos solo para validar y corregir sus propios sistemas.
Rangos afectados
Estado del parche por paquete
| Paquete / producto | Estado | Corregida en | Desde divulgación | Última versión |
|---|---|---|---|---|
pypivllm |
Corregida | 0.31.0 | -22 d | 0.34.0 no afectada |
0.31.0Arsenal público
Sin exploits, módulos ni PoC públicos conocidos a la fecha de verificación.
https://nvd.nist.gov/vuln/detail/cve-2026-114053https://github.com/vllm-project/vllm/security/advisories/GHSA-h3rc-6mm3-gc2mhttps://github.com/vllm-project/vllm/pull/58830https://github.com/vllm-project/vllm/commit/31f2e70cd3e403aaa471fb7fe3d33a7207aaf3cchttps://github.com/vllm-project/vllmhttps://github.com/vllm-project/vllm/releases/tag/v0.31.0https://www.vulncheck.com/advisories/vllm-0.7.3-before-0.31.0-rce-via-mm-processor-kwargs-code-revisionLas referencias externas se muestran para copiar: cve.llama.ec no lo redirige a otros sitios.
