CVE-2026-115913: vllm DoS — Exploit & Fix

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.

Is your site exposed to CVE-2026-115913?

Run a free security scan — no signup, results in seconds.

CVSS v3.1 Score

6.5

MEDIUM

CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:N/I:N/A:H

EPSS — Exploit Prediction

0.0033

Probability of exploitation

0.26%

Percentile rank

EPSS estimates the probability that this vulnerability will be exploited in the wild within the next 30 days. A higher score means more likely to be exploited.

Weakness Type (CWE)

CWE-770 CWE-770

Affected Products

Vendor	Product	Versions
vllm	vllm	0.31.0 — 0.32.0

References

Advisories & Patches

https://github.com/vllm-project/vllm/commit/3d52578e13454de0bb8352cac988347cdcc89928 https://github.com/vllm-project/vllm/pull/60614 https://github.com/vllm-project/vllm/security/advisories/GHSA-8397-8qf6-f2gm

Exploits

https://github.com/vllm-project/vllm/security/advisories/GHSA-8397-8qf6-f2gm

Other References

https://github.com/vllm-project/vllm/releases/tag/v0.32.0

Frequently Asked Questions

What is CVE-2026-115913? +

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. It has a CVSS v3.1 base score of 6.5 (MEDIUM).

How severe is CVE-2026-115913? +

CVE-2026-115913 has a CVSS v3.1 score of 6.5 out of 10, rated MEDIUM. This is a medium-severity vulnerability that should be remediated as part of regular maintenance. The EPSS score is 0.0033, placing it in the 0th percentile for exploitation probability.

What products are affected by CVE-2026-115913? +

CVE-2026-115913 affects products from vllm, specifically: vllm. Check the affected products table above for specific version ranges.

How do I check if I'm vulnerable to CVE-2026-115913? +

You can use Secably's free Website Scanner to check your website for known vulnerabilities. For infrastructure scanning, use the Port Scanner to identify exposed services that may be affected. Check the vendor advisories linked above for specific patch and version information.

Related Vulnerabilities

CVE-2026-39313

mcp-framework is a framework for building Model Context Protocol (MCP) servers. In versions 0.2.21 and below, the readRequestBody() function in …

CVE-2026-35469

spdystream is a Go library for multiplexing streams over SPDY connections. In versions 0.5.0 and below, the SPDY/3 frame parser …

CVE-2026-104817

Misskey is an open source, federated social media platform. From 12.112.1 until 2026.7.0, the unauthenticated fetch-rss API lacks rate limiting …

CVE-2026-75956

Joomla Extension - cmsjunkie.com - DOS vector in pagination parameter handling in J-BusinessDirectory < 6.2.3 - Pagination values were not …

CVE-2026-39803

Allocation of Resources Without Limits or Throttling vulnerability in mtrudel bandit allows unauthenticated remote denial of service via memory exhaustion. …

CVE-2026-39804

Allocation of Resources Without Limits or Throttling vulnerability in mtrudel bandit allows unauthenticated remote denial of service via memory exhaustion …

CVE-2026-115913