CVE-2026-5497

Summary

vLLM versions 0.8.0 and later are vulnerable to an Out-of-Memory (OOM) Denial of Service (DoS) attack due to unbounded frame count processing in the `VideoMediaIO.load_base64()` method. When processing `video/jpeg` data URLs, the method splits the base64 data string on commas to extract individual JPEG frames without enforcing a frame count limit. An attacker can exploit this by crafting a single API request containing thousands of comma-separated base64-encoded JPEG frames in a data URL, causing the server to decode all frames into memory and crash due to excessive memory consumption. This vulnerability is reachable via the OpenAI-compatible chat completions API and does not require authentication.

Impact & exploitability

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

Affected products we track (1)

Recommendation

Apply the vendor fix promptly. Open any affected product above for its exact safe version.

Official patch: https://github.com/vllm-project/vllm/commit/58ee61422169ce17e08248f8efa1e9df434fe395 ↗

Additional information

• NVD record
• https://github.com/vllm-project/vllm/commit/58ee61422169ce17e08248f8efa1e9df434fe395Patch
• https://access.redhat.com/security/cve/CVE-2026-5497
• https://bugzilla.redhat.com/show_bug.cgi?id=2487813
• https://security.access.redhat.com/data/csaf/v2/vex/2026/cve-2026-5497.json
• https://huntr.com/bounties/7bd92629-b396-4449-8f88-6c0092530eb4Advisory