lmdeploy
Affected
- introduced
0.9.2, fixed0.16.0(ecosystem)
Fixed versions
0.16.0
Fix
- Fixed version
0.16.0pip install --upgrade lmdeploy==0.16.0
Advisory
Also known as: CVE-2025-66455, GHSA-2vh9-42vm-xmv2
LMDeploy has Remote Code Execution by Pickle Deserialization via handle_zmq_recv in lmdeploy/lmdeploy/pytorch/disagg/conn/engine_conn.py
Summary
LMDeploy's PyTorch DistServe/PD-disaggregation control plane usedrecv_pyobj() to deserialize messages received through a ZeroMQ PULL
socket. PyZMQ implements recv_pyobj() using Python pickle
deserialization, which can execute arbitrary code while reconstructing
an object.
The peer address used by the receiver was supplied through thePOST /distserve/p2p_connect HTTP endpoint. An attacker who could reach
an affected DistServe API server could cause the server to connect to an
attacker-controlled ZeroMQ endpoint and deserialize a crafted pickle
payload.
API-key authentication is not enabled unless the operator explicitly
configures it. As a result, affected DistServe deployments without API
keys allowed unauthenticated remote code execution with the privileges
of the LMDeploy serving process.
This issue affects the PyTorch backend when PD-disaggregation/DistServe
is enabled. Ordinary deployments that do not use the affected
disaggregated-serving path do not expose this data flow.
Affected components
lmdeploy/serve/openai/endpoints/distserve.py, POST /distserve/p2p_connectDistServeConnectionRequest.remote_engine_endpoint_info.zmq_addresslmdeploy/pytorch/disagg/conn/engine_conn.py, EngineP2PConnection.handle_zmq_recv()recv_pyobj(), which performs pickle deserializationVulnerable data flow
handle_zmq_recv() receives messages using recv_pyobj(). A type check performed after recv_pyobj() cannot mitigate this issue
because pickle payload execution occurs during deserialization.
Impact
Successful exploitation allows arbitrary code execution as the
LMDeploy serving process. This can expose model weights, prompts,
credentials, attached storage, cluster-network services, and host or
GPU resources. An attacker may also modify or terminate the serving
process.
Affected versions
Affected versions:
lmdeploy >= 0.9.2, < 0.16.0The vulnerable P2P receiver was introduced in commit b0b705f7.
Remediation
The issue was fixed by replacing the pickle-based ZeroMQ protocol with
JSON serialization:
send_pyobj() was replaced with send_json().recv_pyobj() was replaced with recv_json().DistServeCacheFreeRequest Pydantic schema before use.Fix commit:
https://github.com/InternLM/lmdeploy/commit/f05b4ad8bf2e2d84101a1d63b3c44fadd99223b2
The fix was released in LMDeploy 0.16.0.
Workarounds
Users who cannot upgrade immediately should:
/distserve/* endpoints.These measures reduce exposure but do not make pickle deserialization
safe. Upgrading to LMDeploy 0.16.0 or later is recommended.
Critical Basis: CVSS v3 score 9.8, vector CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H
0.9.2, fixed 0.16.0 (ecosystem)0.16.0
0.16.0pip install --upgrade lmdeploy==0.16.0Fix commands are the export's own generic commands. Ojuju never runs them. Check them against your setup first.
Check your own computer: guard packages check
Source: PyPA Advisory Database · CC BY 4.0