Skip to main content

LMCache: Remote code execution via unsafe deserialization in multiprocess mode transport (CVE-2026-105192)

0
Critical
Published: 10/16/2026 (10/16/2026, 20:08:44 UTC)
Source: GCVE Database
Product: lmcache

Description

LMCache versions from 0.3.9 up to but not including 0.5.6 have a critical unauthenticated remote code execution vulnerability in multiprocess mode via unsafe pickle deserialization on the ZeroMQ transport port (default 5555). The transport socket, used for inter-process communication, lacks authentication and deserializes attacker-controlled data using pickle.loads, allowing arbitrary code execution as the LMCache process user. By default, the transport binds to localhost, but multi-node deployments often bind to routable addresses, exposing the service. Official container images run the LMCache process as root, increasing risk. The vulnerability is fixed in version 0.5.6 by replacing pickle with a safe, versioned MessagePack format and restricting network exposure.

CVSS v3.1

Score 9.8critical

Attack Vector
Network
Attack Complexity
Low
Privileges Required
None
User Interaction
None
Scope
Unchanged
Confidentiality
High
Integrity
High
Availability
High
CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H

Affected software

PyPIghsa
lmcache
Affected versions
>=0.3.9 <0.5.6

Run on your own infrastructure? Check whether these packages are installed with threat-finder — our free open-source scanner.

AI-Powered Analysis

Machine-generated threat intelligence

AILast updated: 10/16/2026, 23:14:02 UTC

Technical Analysis

LMCache's multiprocess mode opens an unauthenticated ZeroMQ ROUTER socket (default port 5555) for worker processes to share cache blocks. This socket accepts msgpack messages that include an extension type deserialized via pickle.loads without authentication, enabling unauthenticated remote code execution. The vulnerability affects versions from 0.3.9 through 0.5.5 and 0.5.6 release candidates prior to the final 0.5.6 release. The default configuration binds to localhost, but multi-node setups often bind to 0.0.0.0 or other routable addresses, exposing the service to network attackers. The official container images run the LMCache process as root, so exploitation leads to root code execution. The issue is fixed in 0.5.6 by removing pickle deserialization, implementing a typed MessagePack allowlist, changing the wire format, binding example deployments to localhost, and running containers as non-root users.

Potential Impact

An attacker who can connect to the ZeroMQ transport port (default 5555) on a vulnerable LMCache deployment can send a single crafted message that triggers unsafe pickle deserialization, resulting in arbitrary code execution with the privileges of the LMCache process. Since official containers run LMCache as root, this can lead to full system compromise. Single-host default installs binding only to localhost are not exposed remotely, but multi-node or misconfigured deployments binding to routable addresses are vulnerable. The vulnerability has a CVSS 3.1 score of 9.8 (critical).

Mitigation Recommendations

A fixed version 0.5.6 is available and should be deployed. This release replaces pickle deserialization with a safe MessagePack-based format and restricts network exposure by binding example deployments to localhost and running as a non-root user. Until upgrading, do not bind the ZeroMQ transport to routable addresses; restrict access to port 5555 with firewalls or network controls to trusted hosts only.

Pro Console: star threats, build custom feeds, automate alerts via Slack, email & webhooks.Upgrade to Pro

Technical Details

Gcve Source
db.gcve.eu
Osv Id
GHSA-w5f2-9q9c-ff4h
Osv Schema Version
1.4.0
Aliases
["CVE-2026-105192"]
Ecosystems
["PyPI"]
Database Specific Severity
CRITICAL
Cvss Version
3.1

Threat ID: 6ad2aee82cdf04f656455155

Added to database: 10/16/2026, 23:10:32 UTC

Last enriched: 10/16/2026, 23:14:02 UTC

Last updated: 10/16/2026, 23:14:02 UTC

Views: 10

Community Reviews

0 reviews

Crowdsource mitigation strategies, share intel context, and vote on the most helpful responses. Sign in to add your voice and help keep defenders ahead.

Sort by
Loading community insights…

Want to contribute mitigation steps or threat intel context? Sign in or create an account to join the community discussion.

Actions

PRO

Updates to AI analysis require Pro Console access. Upgrade inside Console → Billing.

Please log in to the Console to use AI analysis features.

Need more coverage?

Upgrade to Pro Console for AI refresh and higher limits.

For incident response and remediation, OffSeq services can help resolve threats faster.

Latest Threats

Breach by OffSeqOFFSEQFRIENDS — 25% OFF

Check if your credentials are on the dark web

Instant breach scanning across billions of leaked records. Free tier available.

Scan now
OffSeq TrainingCredly Certified

Lead Pen Test Professional

Technical5-day eLearningPECB Accredited
View courses