CVE-2026-105753 vllm-project CVE debrief

Who should care

Defenders responsible for maintaining and securing systems that use vLLM should prioritize verifying the version of vLLM in use. They should assess exposure to this vulnerability and take steps to remediate it. This includes reviewing compensating controls, monitoring for potential denial-of-service conditions or code execution attempts, and tracking exceptions. Affected operators, platforms, vulnerability-management teams, and security teams should be

Why it matters

A vulnerability in vLLM's multimodal cache can lead to a desync between the frontend and engine core processes, allowing a later request reusing the same media hash to trip a receiver assertion in the engine core. Defenders should prioritize verifying the version of vLLM in use and assessing exposure to this vulnerability.

• Denial-of-service condition due to receiver assertion in the engine core
• Potential for code execution if an attacker can craft a malicious media hash
• Need for verification of vLLM version and exposure to this vulnerability

Technical summary

The vulnerability is caused by a desync between the frontend and engine core processes in vLLM's multimodal cache. When a request is rejected after the frontend has rendered and hashed the multimodal input but before the engine core receives the item, the frontend believes the media is cached while the engine core never received it. A later request reusing the same media hash can then trip a receiver assertion in the engine core.

Defensive priority

Defenders should prioritize verifying the version of vLLM in use and assessing exposure to this vulnerability, as it could lead to a denial-of-service condition or potentially allow for code execution.

Recommended defensive actions

• Verify the version of vLLM in use and update to a fixed version if necessary.
• Assess exposure to this vulnerability and prioritize remediation.
• Monitor for potential denial-of-service conditions or code execution attempts.
• Review compensating controls for exposed systems while remediation is scheduled and verified.
• Check relevant monitoring, detection, and logs for exposed assets that need extra review.
• Track exceptions, retest remediated assets, and close the item only after evidence is documented.
• Confirm whether affected product deployments exist in managed environments and assign an owner for follow-up.

Evidence notes

The vulnerability is caused by a desync between the frontend and engine core processes in vLLM's multimodal cache. When a request is rejected after the frontend has rendered and hashed the multimodal input but before the engine core receives the item, the frontend believes the media is cached while the engine core never received it. A later request reusing the same media hash can then trip a receiver assertion in the engine core.

Sources and references

Methodology and review provenance

AI-assisted synthesis based on stored public vulnerability evidence. System validation, approval state, and publication status do not by themselves establish human review of this revision. PatchSiren helps prioritize defensive review and does not prove exposure or remediation on any system.