Skip to content

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

CriticalCVSS 9.1 EPSS 0.73%
Published
Updated
Exploit status
No public exploit known
CISA SSVC
No known exploitation · automatable: yes · technical impact: total

Description

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.

Upgrade path

3 packages: 2 fixed.

PackageAffectedFixed inStatus
pypi vllm(PyPI) = 0.6.5; = 0.6.6; = 0.6.6.post1; = 0.7.0 +4 more 0.8.0 Fixed
vllm >= 0.6.5, < 0.8.0 0.8.0 Fixed
vllm-project vllm = >= 0.6.5, < 0.8.0 No fix listed Versions listed

Affected products

ProductAffected versionsFixed in
pypi vllm(PyPI) >= 0.6.5, < 0.8.0 0.8.0
pypi vllm(PyPI) = 0.6.5 No fix listed
pypi vllm(PyPI) = 0.6.6 No fix listed
pypi vllm(PyPI) = 0.6.6.post1 No fix listed
pypi vllm(PyPI) = 0.7.0 No fix listed
pypi vllm(PyPI) = 0.7.1 No fix listed
pypi vllm(PyPI) = 0.7.2 No fix listed
pypi vllm(PyPI) = 0.7.3 No fix listed
vllm-project vllm = >= 0.6.5, < 0.8.0 No fix listed
vllm >= 0.6.5, < 0.8.0 0.8.0

References