vLLM Cohere2 Vision denial of service via max_patches
Key facts
- Attack vector
- Network
- Privileges required
- None
- User interaction
- None
- Attack complexity
- Low
- Attack requirements
- None
- Impact type
- Denial of service
- Confidentiality · integrity · availability
- None · None · High
- Fix status
- Fixed
Attack profile from CVSS 4.0 · CNA. Fix status from CNA. Impact type from CWE-400.
Summary
vLLM from 0.10.1 before 0.31.0 passes a request-controlled mm_processor_kwargs.max_patches value to the Hugging Face processor for Cohere2 Vision models without an upper bound. This causes quadratic CPU work during multimodal preprocessing (CWE-400). A remote client can send small image chat-completion requests with large, varying max_patches values, which bypass the LRU cache and exhaust preprocessing workers, denying service to other clients.
Written with Claude Haiku 5.5 from the sources below · 2026-11-13
Description from the CNA
vLLM from 0.10.1 before 0.31.0 forwards a request-controlled mm_processor_kwargs.max_patches value to the Hugging Face processor for Cohere2 Vision models without a server-side upper bound, causing quadratic CPU work during multimodal preprocessing. Remote clients can send small image chat-completion requests with large, varying max_patches values to bypass the LRU cache, exhaust preprocessing workers, and deny service to other clients.
Why it matters now
SSVC decision points
Not available: CISA has not published SSVC decision points for this record.
Affected products
| Vendor | Product | Affected versions | Fixed in | Status | Source |
|---|---|---|---|---|---|
| vllm-project | vllm | 0.10.1 to before 0.31.0 | 0.31.0 | Patch available | CNA |
- vllm Vendorvllm-project Affected 0.10.1 to before 0.31.0 Fixed in 0.31.0 Patch available CNA
Remediation
Upgrade vLLM to 0.31.0. The sources do not list a GHSA or advisory fix reference beyond the vLLM release that contains the fix.
Written with Claude Haiku 5.5 from the sources below · 2026-11-13
- Fix · CNA vllm (vllm-project) Fixed in 0.31.0
Fixed in the same release
Records fixed by the same vendor advisory or release, as the CNA lists them.
- Release ·
vllm-project vllm 0.31.0
5 records Highest priority Priority LOW 5 LOW
Other records, highest priority first
- TI-2026-91289 Resource exhaustion in vLLM MiniMax M3 via min_pixelsPriority LOW 37
- TI-2026-91288 Information disclosure in vLLM sanitize_message() leaks heap addressesPriority LOW 36
- TI-2026-91292 Resource exhaustion in vLLM via request-level num_framesPriority LOW 31
Timeline
- 2026-11-13 Threatist record created from CVE-2026-126397
- 2026-11-13 Alias added: GHSA-h8w4-vgpr-474w
- 2026-11-14 EPSS moved to the below the 50th percentile
References
Patch 1
More vllm-project vulnerabilities
- TI-2024-10229 Remote Code Execution in vllm-project/vllm CVSS 3.0CRITICAL 9.8
- TI-2024-08404 Remote Code Execution in vllm-project/vllm CVSS 3.1CRITICAL 9.8
- TI-2026-36017 vLLM: OpenAI auth bypass CVSS 3.1CRITICAL 9.1
- TI-2026-63056 vLLM before 0.28.0 Denial of Service via negative token ID CVSS 4.0HIGH 8.7
- TI-2026-04308 Hardcoded trust_remote_code=True in vllm-project/vllm Bypasses User Security Control CVSS 3.0HIGH 8.8
- TI-2026-39137 vLLM: Remote DoS in vLLM via Invalid Recovered Token Reinjection CVSS 3.1HIGH 7.5
Sources
- CVE-2026-126397 · CVE Program
- CVE-2026-126397 · NVD
- FIRST EPSS
- GHSA-h8w4-vgpr-474w · GitHub Advisory Database · CC BY 4.0
This product uses the NVD API but is not endorsed or certified by the NVD. Sources and licences