CVE-2025-29783: vLLM allows remote code execution via Mooncake integration
vLLM
Unsafe pickle deserialization exposed over ZMQ and TCP in vLLM when configured with Mooncake allows remote code execution. A patch and advisory were issued by maintainers.
disclosed CVENo confirmed exploitation, so this record is tracked as a disclosed vulnerability and not counted as an incident. It moves into the counts if CISA or the vendor confirms exploitation. Why
- CVSS
- 9.0
- EPSS
- 0.01 · top 45% · 2026-11-13
- fix status
- Fix status unknown
- first report
- 19 Mar 2025
- last update
- 19 Mar 2025
- CVE ids
Timeline · 2
Sources · 2
Advisories and databases
- 2025-03-19CVE-2025-29783: vLLM Allows Remote Code Execution via Mooncake IntegrationGitHub Advisory Database · github.com/advisories/GHSA-x3m8-f7g5-qhm7
- 2025-03-19CVE-2025-29783: vLLM is a high-throughput and memory-efficient inference and serving engine for LLMs.NVD · nvd.nist.gov/vuln/detail/CVE-2025-29783
Titles and links belong to their publishers. The summary above is written by gcdRadar in its own words; data is CC BY-SA 4.0.
Classification
- lane
- Patch now
- AI role
- Attack on AI
- categories
- Infrastructure vulnerability
- severity
- 2 · Disclosed vulnerability
- fix status
- Fix status unknown
- targets
- Framework / tooling
- owasp llm top 10
- LLM03:2025 Supply Chain
- mitre atlas
- no technique for these categories
- dates
- happened 19 Mar 2025 · first reported 19 Mar 2025
- origin
- GitHub Advisory DatabaseVulnerability record from NVD / the GitHub Advisory Database. This product uses the NVD API but is not endorsed or certified by the NVD.
- record
- confidence 0.95 · created 2026-10-18 07:00Z · updated 2026-10-18 10:04Z
Lane, AI role, categories and severity come from the classifier; the OWASP and ATLAS rows are derived from the categories. How these are assigned
Copy a brief
This record as a short Markdown brief, for a ticket, a chat or a coding agent.
the brief as text
# CVE-2025-29783: vLLM allows remote code execution via Mooncake integration https://gcdradar.vercel.app/incidents/cve-2025-29783-vllm-allows-remote-code-execution-via-mooncak · AIR-2025-0206 · closed 2025-03-19 - Lane: Patch now - Severity: 2 (Disclosed vulnerability) - Fix status: Fix status unknown - First reported: 2025-03-19 - AI role: Attack on AI - Products: vLLM - CVE ids: CVE-2025-29783 - CVSS: 9 ## Summary Unsafe pickle deserialization exposed over ZMQ and TCP in vLLM when configured with Mooncake allows remote code execution. A patch and advisory were issued by maintainers. ## What to do No action recorded; read the sources below and the vendor's advisory. ## Latest updates - 2025-03-19: CVE-2025-29783: vLLM is a high-throughput and memory-efficient inference and serving engine for LLMs. (https://nvd.nist.gov/vuln/detail/CVE-2025-29783) - 2025-03-19: CVE-2025-29783: vLLM Allows Remote Code Execution via Mooncake Integration (https://github.com/advisories/GHSA-x3m8-f7g5-qhm7) ## Sources - 2025-03-19 GitHub Advisory Database: CVE-2025-29783: vLLM Allows Remote Code Execution via Mooncake Integration https://github.com/advisories/GHSA-x3m8-f7g5-qhm7 - 2025-03-19 NVD: CVE-2025-29783: vLLM is a high-throughput and memory-efficient inference and serving engine for LLMs. https://nvd.nist.gov/vuln/detail/CVE-2025-29783 Verify against the vendor's advisory before changing production. Data CC BY-SA 4.0, https://gcdradar.vercel.app.
Related
- closedvLLM vLLM Vulnerable to Denial of Service by Abusing XGrammar Cache
Patch now · Disclosed · since 15 Apr 2025
AI - closed9.8vLLM CVE-2024-11041: vLLM Deserialization of Untrusted Data vulnerability
Patch now · Disclosed · since 20 Mar 2025 · CVE-2024-11041
AI - closed9.8vLLM CVE-2024-9053: vLLM Pickle Deserialization Remote Code Execution
Patch now · Disclosed · since 20 Mar 2025 · CVE-2024-9053
AI - closed9.8vLLM CVE-2024-9052: vLLM Deserialization Vulnerability in GroupCoordinator
Patch now · Disclosed · since 20 Mar 2025 · CVE-2024-9052
AI - closed6.5vLLM CVE-2025-29770: vLLM denial of service via outlines unbounded cache
Patch now · Disclosed · since 19 Mar 2025 · CVE-2025-29770
AI - closed7.5vLLM CVE-2025-48956: vLLM API Denial of Service via Large HTTP Headers
Patch now · Disclosed · since 21 Aug 2025 · CVE-2025-48956
AI