Vulnerability Description
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.
CVSS Score
MEDIUM
Affected Products
| Vendor | Product | Versions |
|---|---|---|
| Vllm | Vllm | >= 0.31.0, < 0.32.0 |
Related Weaknesses (CWE)
References
- https://github.com/vllm-project/vllm/commit/3d52578e13454de0bb8352cac988347cdcc8Patch
- https://github.com/vllm-project/vllm/pull/60614Patch
- https://github.com/vllm-project/vllm/releases/tag/v0.32.0Release Notes
- https://github.com/vllm-project/vllm/security/advisories/GHSA-8397-8qf6-f2gmExploitPatchVendor Advisory
FAQ
What is CVE-2026-115913?
CVE-2026-115913 is a vulnerability with a CVSS score of 6.5 (MEDIUM). 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 ...
How severe is CVE-2026-115913?
CVE-2026-115913 has been rated MEDIUM with a CVSS base score of 6.5/10. Review the CVSS metrics above for detailed severity breakdown.
Is there a patch for CVE-2026-115913?
Check the references section above for vendor advisories and patch information. Affected products include: Vllm Vllm.