Authentication Bypass and Privilege Escalation Vulnerability in OpenShift AI

CVE-2024-7557

Summary

In OpenShift AI, a vulnerability exists that permits authentication bypass and privilege escalation across models within the same namespace. This concern arises during AI model deployment, where the user interface allows the configuration of authentication for models. Unfortunately, tokens from one model are inadvertently usable for accessing other models and APIs within the same namespace. The vulnerability is exacerbated by the exposure of ServiceAccount tokens in the UI, enabling malicious users to exploit these tokens through commands like 'oc --token={token}', thereby gaining unauthorized access to resources and elevating privileges significantly.

References