CVE-2026-68903
Received Received - Intake

Arbitrary File Read and Host Path Mount in LLMKube Kubernetes Operator

Vulnerability report for CVE-2026-68903, including description, CVSS score, EPSS score, affected products, exploitability, helpful resources, and attack-flow context.

Publication date: 2026-11-18

Last updated on: 2026-11-18

Assigner: GitHub, Inc.

Description

LLMKube is a Kubernetes operator for self-hosted large language model inference across heterogeneous hardware. Prior to 0.9.0, the ModelReconciler accepts arbitrary absolute paths and file:// URIs in Model.spec.source without restricting them to approved roots. A namespace user who can create or update Model and InferenceService resources can cause copyLocalModel to open an attacker-selected file in the controller pod or cause a HostPathVolumeSource to mount an attacker-selected node file when modelCache.enabled is enabled, which is the default. The selected file is copied into model storage and can expose controller service-account credentials or other node secrets outside the attacker's namespace. This issue is fixed in version 0.9.0.

CVSS Scores

EPSS Scores

Probability:
Percentile:

Meta Information

Published
2026-11-18
Last Modified
2026-11-18
Generated
2026-11-19
AI Q&A
2026-11-19
EPSS Evaluated
N/A
NVD
EUVD

Affected Vendors & Products

Showing 3 associated CPEs
Vendor Product Version / Range
llmkube llmkube to 0.9.0 (inc)
defilantech llmkube to 0.9.0 (exc)
defilantech llmkube 0.9.0

Helpful Resources

Exploitability

CWE
CWE Icon
KEV
KEV Icon
CWE ID Description
CWE-22 The product uses external input to construct a pathname that is intended to identify a file or directory that is located underneath a restricted parent directory, but the product does not properly neutralize special elements within the pathname that can cause the pathname to resolve to a location that is outside of the restricted directory.

Attack-Flow Graph

AI Quick Actions

Instant insights powered by AI
Executive Summary

CVE-2026-68903 is a high-severity vulnerability in LLMKube (versions <= 0.8.28) that allows arbitrary host file read via path injection in the Model.spec.source field. The ModelReconciler accepts absolute paths or file:// URIs without validation, enabling two attack vectors: (1) reading arbitrary files in the controller pod via os.Open() and (2) mounting node files into inference pods via HostPathVolumeSource. Model caching, enabled by default, triggers these sinks.

Detection Guidance

Check LLMKube version with kubectl get deployment -n <namespace> <llmkube-deployment> -o jsonpath='{.spec.template.spec.containers[0].image}' to see if it is <= 0.8.28. Inspect Model and InferenceService resources for absolute paths or file:// URIs in Model.spec.source fields using kubectl get model -A -o yaml or kubectl get inferenceservice -A -o yaml.

Impact Analysis

An attacker with namespace-level permissions to create/update Model or InferenceService resources can exploit this to read sensitive files like Kubernetes service account tokens or node secrets (e.g., /etc/passwd, /etc/shadow). This can lead to privilege escalation, lateral movement across the cluster, or exposure of confidential data.

Compliance Impact

This vulnerability could lead to unauthorized access to sensitive data, violating GDPR (data protection) and HIPAA (health information privacy) by exposing personal or protected health information. Compliance requires preventing such breaches through secure configurations and patching.

Mitigation Strategies

Upgrade LLMKube to version 0.9.0 or later immediately. Configure allowlists for modelSource.allowedHostPathRoots and modelSource.allowedRemoteHosts in Helm values or via controller flags. Delete any existing InferenceServices with sources outside permitted ranges.

Chat Assistant

Ask questions about this CVE
Hi! I’m here to help you understand CVE-2026-68903. Ask me anything about the vulnerability, its impact, or mitigation strategies.
0/70

EPSS Chart

Security Radar

News mentioning this vulnerability
Loading related security news...