AVID-2026-R1678

Description

vLLM Allows Remote Code Execution via Mooncake Integration (CVE-2025-29783)

Details

vLLM is a high-throughput and memory-efficient inference and serving engine for LLMs. When vLLM is configured to use Mooncake, unsafe deserialization exposed directly over ZMQ/TCP on all network interfaces will allow attackers to execute remote code on distributed hosts. This is a remote code execution vulnerability impacting any deployments using Mooncake to distribute KV across distributed hosts. This vulnerability is fixed in 0.8.0.

Reason for inclusion in AVID: The CVE-2025-29783 entry describes a remote code execution vulnerability in vLLM caused by unsafe deserialization over Mooncake integration via ZMQ/TCP. This directly concerns AI software stacks (model serving/deployment) and affects components used to build/run general-purpose AI systems. It is a security vulnerability (RCE) with explicit exploitation potential and CVE/NVD references provide evidence. Therefore it meets all AVID criteria for inclusion in AI supply chain curation.

References

• NVD entry
• https://github.com/vllm-project/vllm/security/advisories/GHSA-x3m8-f7g5-qhm7
• https://github.com/vllm-project/vllm/pull/14228
• https://github.com/vllm-project/vllm/commit/288ca110f68d23909728627d3100e5a8db820aa2

Affected or Relevant Artifacts

• Developer: vllm-project
• Deployer: vllm-project
• Artifact Details:

Impact

AVID Taxonomy Categorization

• Risk domains: Security
• SEP subcategories: S0100: Software Vulnerability
• Lifecycle stages: L06: Deployment

CVSS

CWE

Other information

• Report Type: Advisory
• Credits:
• Date Reported: 2025-03-19
• Version: 0.3.3
• AVID Entry