Improper Input Validation in LLMKube Kubernetes Operator
CVE-2026-68903

7.7HIGH

Key Information:

Status
Vendor
CVE Published:
18 November 2026

What is CVE-2026-68903?

LLMKube, a Kubernetes operator facilitating large language model inference, is susceptible to a vulnerability where it allows arbitrary absolute paths and file:// URIs in Model.spec.source. This oversight permits namespace users to manipulate Model and InferenceService resources, thereby compromising security by opening attacker-selected files or mounting node files within the controller pod. This can lead to exposure of sensitive information such as service-account credentials or node secrets, particularly when modelCache.enabled is activated by default. The weakness was addressed in version 0.9.0, reinforcing the importance of strict input validation and resource handling.

Affected Version(s)

LLMKube < 0.9.0

References

CVSS V3.1

Score:
7.7
Severity:
HIGH
Confidentiality:
High
Integrity:
None
Availability:
High
Attack Vector:
Network
Attack Complexity:
Low
Privileges Required:
Low
User Interaction:
None
Scope:
Changed

Timeline

  • Vulnerability published

  • Vulnerability Reserved

.