vLLM is an inference and serving engine for large language models (LLMs). From version 0.5.5 to before 0.11.1, users can crash the vLLM engine serving multimodal models by passing multimodal embedding inputs with correct ndim but incorrect shape (e.g. hidden dimension is wrong), regardless of whether the model is intended to support such inputs (as defined in the Supported Models page). This issue has been patched in version 0.11.1.
Affected: PyPI: vllm
### Summary
Users can crash the vLLM engine serving multimodal models by passing multimodal embedding inputs with correct `ndim` but incorrect `shape` (e.g. hidden dimension is wrong), regardless of whether the model is intended to support such inputs (as defined in the Supported Models page).
The issue has existed ever since we added support for image embedding inputs, i.e. #6613 (released in v0.5.5)
### Details
Using image embeddings as an example:
- For models that support image embedding inputs, the engine crashes when scattering the embeddings to `inputs_embeds` (mismatched shape) - For models that don't support image embedding inputs, the engine crashes when validating the inputs inside `get_input_embeddings` (validation fails).
This happens because we only validate `ndim` of the tensor, but not the full shape, in input processor (via `MultiModalDataParser`).
### Impact
- Denial of service by crashing the engine
### Mitigation
- Use API key to limit access to trusted users. - Set `--limit-mm-per-prompt` to 0 for all non-text modalities to ban multimodal inputs, which includes multimodal embedding inputs. However, the model would then only accept text, defeating the purpose of using a multi-modal model.
### Resolution
- https://github.com/vllm-project/vllm/pull/27204
Affected: PyPI: vllm
### Summary
Users can crash the vLLM engine serving multimodal models by passing multimodal embedding inputs with correct `ndim` but incorrect `shape` (e.g. hidden dimension is wrong), regardless of whether the model is intended to support such inputs (as defined in the Supported Models page).
The issue has existed ever since we added support for image embedding inputs, i.e. #6613 (released in v0.5.5)
### Details
Using image embeddings as an example:
- For models that support image embedding inputs, the engine crashes when scattering the embeddings to `inputs_embeds` (mismatched shape) - For models that don't support image embedding inputs, the engine crashes when validating the inputs inside `get_input_embeddings` (validation fails).
This happens because we only validate `ndim` of the tensor, but not the full shape, in input processor (via `MultiModalDataParser`).
### Impact
- Denial of service by crashing the engine
### Mitigation
- Use API key to limit access to trusted users. - Set `--limit-mm-per-prompt` to 0 for all non-text modalities to ban multimodal inputs, which includes multimodal embedding inputs. However, the model would then only accept text, defeating the purpose of using a multi-modal model.
### Resolution
- https://github.com/vllm-project/vllm/pull/27204
openssl: Out-of-bounds read & write in RFC 3211 KEK Unwrap libxslt: libxml2: Inifinite recursion at exsltDynMapFunction function in libexslt/dynamic.c golang.org/x/oauth2/jws: Unexpected memory consumption during token parsing in golang.org/x/oauth2/jws golang.org/x/crypto/ssh: Denial of Service in the Key Exchange of golang.org/x/crypto/ssh runc: container escape with malicious config due to /dev/console mount and related races firefox: thunderbird: expat: libexpat in Expat allows attackers to trigger large dynamic memory allocations via a small document that is submitted for parsing vllm: VLLM deserialization vulnerability leading to DoS and potential RCE vllm: vLLM vulnerable to DoS with incorrect shape of multimodal embedding inputs ray: Ray is vulnerable to RCE via Safari & Firefox Browsers through DNS Rebinding Attack vllm: vLLM: Remote Code Execution via malicious model configuration github.com/sigstore/fulcio: Fulcio: Denial of Service via crafted OpenID Connect (OIDC) token
Ein entfernter, authentisierter Angreifer kann mehrere Schwachstellen in vllm ausnutzen, um einen Denial of Service Angriff durchzuführen oder Code auszuführen.