A critical vulnerability in OpenML.org version v2.0.20241110 allows unauthenticated attackers to hijack user accounts by brute-forcing predictable security tokens.

The application utilizes predictable MD5-based tokens for sensitive workflows, including password resets and account confirmations, which are generated using only non-random timestamps. This flaw allows an unauthenticated remote attacker to guess valid tokens and perform unauthorized account actions.

The ability for an attacker to hijack user accounts poses a significant risk to data integrity and user privacy. Successful exploitation could lead to unauthorized access to research data, account takeover, and potential loss of intellectual property, justifying the high severity of this vulnerability.

Immediate Action: Review the official OpenML security advisory for available patches or configuration changes and apply them immediately to replace the insecure token generation mechanism.

Proactive Monitoring: Monitor authentication logs for an unusual volume of token verification failures or rapid sequences of account confirmation attempts from single or distributed IP sources.

Compensating Controls: Implement rate limiting on all endpoints associated with password resets, email changes, and account verification to impede brute-force attempts.

Public Exploit Available: Unknown.

Given the potential for full account takeover, this vulnerability represents a high risk to the OpenML platform. Administrators must prioritize the application of vendor-provided patches or security updates. If a direct patch is not yet available, implement strict rate limiting and monitor logs for signs of brute-force activity to mitigate the risk of unauthorized access.