CVE-2026-56340 - Exploits & Severity

vLLM versions 0.10.2 through 0.12.x are missing sparse tensor validation in multimodal embeddings processing. When the prompt-embeds feature is enabled, attackers can submit crafted embedding requests with malformed tensor indices (negative or out-of-bounds values) to trigger crashes, resource exhaustion, or out-of-bounds memory corruption.

An authenticated user over the network with the ability to submit embedding requests can cause denial of service through crashes and resource exhaustion, or potentially trigger out-of-bounds memory corruption that could lead to arbitrary code execution.

There is no evidence that a public proof-of-concept exists. There is no evidence of proof of exploitation at the moment.

Upgrade vLLM to version 0.13.0 or later. Disable the prompt-embeds feature if upgrading is not immediately possible. Restrict network access to embedding endpoints to authenticated users only.

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

CVSS:4.0/AV:N/AC:L/AT:N/PR:L/UI:N/VC:H/VI:H/VA:H/SC:N/SI:N/SA:N/E:X/CR:X/IR:X/AR:X/MAV:X/MAC:X/MAT:X/MPR:X/MUI:X/MVC:X/MVI:X/MVA:X/MSC:X/MSI:X/MSA:X/S:X/AU:X/R:X/V:X/RE:X/U:X

NVD published the first details for CVE-2026-56340

Feedly found the first article mentioning CVE-2026-56340 . See article

GitHub Advisories released a security advisory .

CVE-2026-56340 - Exploits & Severity - Feedly

CVE-2026-56340: vLLM - Denial of Service via Unvalidated Multimodal Embeddings [HIGH] CVSS 8.8

CVE-2026-56340 vLLM versions >= 0.10.2 and < 0.13.0 are missing sparse tensor validation in multimodal embeddings processing. Because PyTorch disables sparse tensor invariant checks…

CVE-2026-56340 - vLLM - Denial of Service via Unvalidated Multimodal Embeddings CVE ID : CVE-2026-56340 Published : June 20, 2026, 6:27 p.m. | 1 hour, 16 minutes ago Description : vLLM versions &gt;= 0.10.2 and &lt; 0.13.0 are missing sparse tensor validation in multimod...

Collect, analyze, and vulnerability reports faster using AI