← Vulnerability feed

Vulnerability record · CVE-2025-47277 · published 20 May 2025

CVE-2025-47277: Vllm deserialization of untrusted data vulnerability

Vllm · Vllm

vLLM, an inference and serving engine for large language models (LLMs), has an issue in versions 0.6.5 through 0.8.4 that ONLY impacts environments using the `PyNcclPipe` KV cache transfer integration with the V0 engine. No other configurations are affected. vLLM supports the use of the `PyNcclPipe` class to establish a peer-to-peer communication domain for data transmission between distributed nodes. The GPU-side KV-Cache transmission is implemented through the `PyNcclCommunicator` class, while CPU-side control message passing is handled via the `send_obj` and `recv_obj` methods on the CPU side.​ The intention was that this interface should only be exposed to a private network using the IP address specified by the `--kv-ip` CLI parameter. The vLLM documentation covers how this must be limited to a secured network. The default and intentional behavior from PyTorch is that the `TCPStore` interface listens on ALL interfaces, regardless of what IP address is provided. The IP address given was only used as a client-side address to use. vLLM was fixed to use a workaround to force the `TCPStore` instance to bind its socket to a specified private interface. As of version 0.8.5, vLLM limits the `TCPStore` socket to the private interface as configured.

9.8 CVSS 3.1 Critical EPSS 0.96% · top 40.0% CWE-502 · Deserialization of untrusted data
9.8CVSS 3.1 base score
0.96%EPSS exploitation probability, 30 days
NoNot in CISA KEV
1Affected product versions listed by NVD
4References, 1 tagged exploit
17 Jun 2026Last modified by NVD

Description

vLLM, an inference and serving engine for large language models (LLMs), has an issue in versions 0.6.5 through 0.8.4 that ONLY impacts environments using the `PyNcclPipe` KV cache transfer integration with the V0 engine. No other configurations are affected. vLLM supports the use of the `PyNcclPipe` class to establish a peer-to-peer communication domain for data transmission between distributed nodes. The GPU-side KV-Cache transmission is implemented through the `PyNcclCommunicator` class, while CPU-side control message passing is handled via the `send_obj` and `recv_obj` methods on the CPU side.​ The intention was that this interface should only be exposed to a private network using the IP address specified by the `--kv-ip` CLI parameter. The vLLM documentation covers how this must be limited to a secured network. The default and intentional behavior from PyTorch is that the `TCPStore` interface listens on ALL interfaces, regardless of what IP address is provided. The IP address given was only used as a client-side address to use. vLLM was fixed to use a workaround to force the `TCPStore` instance to bind its socket to a specified private interface. As of version 0.8.5, vLLM limits the `TCPStore` socket to the private interface as configured.

CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H

Affected products

1 vulnerable configurations from NVD's CPE data, grouped by vendor and product.

References

LinkTags
https://docs.vllm.ai/en/latest/deployment/security.html Technical Description
https://github.com/vllm-project/vllm/commit/0d6e187e88874c39cda7409cf673f9e6546893e7 Patch
https://github.com/vllm-project/vllm/pull/15988 Issue TrackingPatch
https://github.com/vllm-project/vllm/security/advisories/GHSA-hjq4-87xh-g4fv ExploitVendor Advisory

Track CVE-2025-47277 inside VULONE

Watch it alongside the ransomware crews, C2 infrastructure and forum chatter that reference it, query it through the API and pull it into your SIEM over TAXII.

Start free Open in platform

Related vulnerabilities

Same products first, then exploited flaws of the same weakness class.

9.8CVE-2026-25960Vllm server-side request forgery (ssrf) vulnerabilityvLLM is an inference and serving engine for large language models (LLMs). The SSRF protection fix for CVE-2026-24779 add in 0.15.1 can be bypassed in…EPSS 0.72%9.8CVE-2026-22778Vllm sensitive information in log file vulnerabilityvLLM is an inference and serving engine for large language models (LLMs). From 0.8.3 to before 0.14.1, when an invalid image is sent to vLLM's multim…EPSS 10%9.8CVE-2026-22807Vllm code injection vulnerabilityvLLM is an inference and serving engine for large language models (LLMs). Starting in version 0.10.1 and prior to version 0.14.0, vLLM loads Hugging …EPSS 0.83%9.8CVE-2025-32444Vllm deserialization of untrusted data vulnerabilityvLLM is a high-throughput and memory-efficient inference and serving engine for LLMs. Versions starting from 0.6.5 and prior to 0.8.5, having vLLM in…EPSS 1.8%9.8CVE-2024-11041Vllm deserialization of untrusted data vulnerabilityvllm-project vllm version v0.6.2 contains a vulnerability in the MessageQueue.dequeue() API function. The function uses pickle.loads to parse receive…EPSS 1.6%9.1CVE-2026-48746Vllm http request smuggling vulnerabilityvLLM is an inference and serving engine for large language models (LLMs). From 0.3.0 until 0.22.0, a vulnerability in ASGI web servers and starlette'…EPSS 1.2%9.0CVE-2025-29783Vllm deserialization of untrusted data vulnerabilityvLLM is a high-throughput and memory-efficient inference and serving engine for LLMs. When vLLM is configured to use Mooncake, unsafe deserialization…EPSS 0.73%8.8CVE-2026-54232Vllm uncontrolled search path element vulnerabilityvLLM is an inference and serving engine for large language models (LLMs). Prior to 0.22.1, the vLLM Dockerfile is vulnerable to a dependency confusio…EPSS 0.56%

Source: NIST National Vulnerability Database (record CVE-2025-47277), CISA KEV, FIRST EPSS (scores of 2026-09-27). This page is refreshed as NVD updates the record.