CVE-2026-37237 vLLM Project CVE debrief

Who should care

Defenders and administrators of systems using the vLLM project, particularly those using versions up to and including 0.17.0, should assess their exposure and take measures to prevent potential DoS attacks.

Why it matters

CVE-2026-37237 is a high-severity vulnerability in the vLLM project that allows remote attackers to cause a Denial of Service via memory exhaustion. Defenders and administrators of systems using the vLLM project should assess their exposure and take measures to prevent potential DoS attacks.

• Potential Denial of Service (DoS) via memory exhaustion.
• Need to verify vLLM version and apply patches to prevent exploitation.
• Possible impact on server availability and performance.

Technical summary

The AsyncMediaIO.fetch_audio and AsyncMediaIO.fetch_image functions in multimodal/inputs.py fetch user-supplied media URLs using aiohttp and call r.read() without enforcing a maximum response size, allowing an attacker to exhaust server memory by providing a URL to an arbitrarily large file. This vulnerability affects vLLM versions up to and including 0.17.0, and defenders should prioritize verifying the vLLM version and applying patches to prevent potential DoS attacks. The lack of a maximum response size limit allows an attacker to cause a Denial of Service via memory exhaustion.

Defensive priority

Defenders should prioritize verifying the vLLM version and applying patches to prevent potential DoS attacks.

Recommended defensive actions

• Verify the vLLM version and check if it is within the affected range (up to and including 0.17.0).
• Apply patches or updates to fix the vulnerability.
• Implement input validation and size limits for user-supplied media URLs.
• Monitor server memory usage and implement measures to prevent memory exhaustion.
• 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.

Evidence notes

The CVE record and NVD vulnerability detail page provide information on the vulnerability, including its description, CVSS score, and affected versions. The AsyncMediaIO.fetch_audio and AsyncMediaIO.fetch_image functions in multimodal/inputs.py fetch user-supplied media URLs using aiohttp and call r.read() without enforcing a maximum response size, allowing an attacker to exhaust server memory by providing a URL to an arbitrarily large file. Defenders should verify the vLLM version and check if it is within the affected range (up to 0

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.