medium6.5
CVE-2026-115913
vLLM: Phi-4 multimodal audio encoder materializes and retains full attention matrices: a single max-length audio OOM-crashes the vLLM engine
Impact
vLLM is an inference and serving engine for large language models. Prior to 0.32.0, the Phi-4 multimodal audio encoder in phi4mm_audio.py and phi4mm_utils.py uses eager attention by default, allowing a remote client to submit one legal long audio clip that causes MultiHeadedAttention.forward_attention to materialize and retain full attention matrices in self.attn. The scaled dot-product attention branch does not materialize these matrices. The resulting quadratic GPU-memory use can cause an out-of-memory condition and terminate EngineCore for all users. This issue is fixed in version 0.32.0.
Affected packages
| Ecosystem | Package | Affected | Fixed in |
|---|---|---|---|
| pip | vllm | < 0.32.0 | 0.32.0 |
Affected products
| Vendor | Product | Affected versions |
|---|---|---|
| Vllm | Vllm | from 0.31.0 before 0.32.0 |
Weakness classification
References
- NVD record
- CVE.org record
- GitHub advisory
- https://github.com/advisories/GHSA-8397-8qf6-f2gm
- https://github.com/vllm-project/vllm/security/advisories/GHSA-8397-8qf6-f2gm
- https://nvd.nist.gov/vuln/detail/CVE-2026-115913
- https://github.com/vllm-project/vllm/pull/60614
- https://github.com/vllm-project/vllm/commit/3d52578e13454de0bb8352cac988347cdcc89928
- https://github.com/vllm-project/vllm/releases/tag/v0.32.0