CVE-2026-115913
What it is
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.
At a glance
- Exploitability
- Can be exploited over the network with a low-privilege account. High risk of service disruption.
- Exploit odds
- 0.3% chance of exploitation in the next 30 days (EPSS).
- Fix
- A patch, advisory or fixed version is referenced. Upgrade or apply it.
Timeline
- ReservedThe CVE id was reserved by its CNA
- PublishedDisclosed publicly on CVE.org and NVD
- Last updatedMost recent change to the record
Why it ranks here
- Exploited in the wild: no, CISA does not list it.
- Exploit odds: 0.3% in the next 30 days (EPSS).
- Severity: 6.5 MEDIUM (CVSS).
- Fix: a patch or advisory is linked.
Details
- Published
- Last modified
- Exploit probability
- 0.3% · EPSS, next 30 days
- Weakness
- CWE-770 · Allocation without limits
- Record status
- Analyzed
Affected
- Vllm < 0.32.0
Attack profile CVSS 3.1 · 6.5
- Attack vector
- Network
- Complexity
- Low
- Privileges needed
- Low
- User interaction
- None
- Scope
- Unchanged
- Confidentiality
- None
- Integrity
- None
- Availability
- High
Raw vector
CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:N/I:N/A:H
Is it in your code?
Hand this to your coding agent. It searches your project for the affected products and checks versions. Nothing leaves your machine.
Read https://cve-list.com/agent/cve/CVE-2026-115913.md and follow it. Check whether this project is affected by CVE-2026-115913. Read only. Don't change anything and don't upload my code. Tell me yes, no or not sure, with the file as proof.
Fix and advisories
- https://github.com/vllm-project/vllm/commit/3d52578e13454de0bb8352cac988347cdcc89928
- https://github.com/vllm-project/vllm/pull/60614
- https://github.com/vllm-project/vllm/releases/tag/v0.32.0
- https://github.com/vllm-project/vllm/security/advisories/GHSA-8397-8qf6-f2gm