A critical deserialization vulnerability in the MONAI AI toolkit allows authenticated attackers to achieve remote code execution by loading malicious model checkpoints.

This vulnerability is classified as CWE-502: Deserialization of Untrusted Data, where the software insecurely loads model checkpoints. The vulnerability allows an attacker with low privileges to trigger arbitrary code execution by providing crafted malicious content during the deserialization process.

Successful exploitation of this vulnerability allows for remote code execution on the underlying host system, potentially leading to a total compromise of the medical imaging environment. Given the high CVSS score of 8.8, this flaw poses a severe threat to data integrity and system availability, which could disrupt critical healthcare operations and expose sensitive patient data.

Immediate Action: Update the MONAI package to version 1.5.1 or later via PyPI to incorporate the necessary security fixes for checkpoint loading.

Proactive Monitoring: Review application logs for unusual model loading activity or attempts to deserialize objects from untrusted or external sources.

Compensating Controls: Implement strict network segmentation to ensure that model loading processes are isolated from external networks and strictly control the source of all pre-trained model checkpoints.

Public Exploit Available: Unknown.

The vulnerability represents a significant risk to the integrity of AI-driven healthcare infrastructure. Organizations currently utilizing MONAI versions 1.5.0 or earlier must prioritize upgrading to version 1.5.1 immediately to eliminate the risk of arbitrary code execution. If an immediate update is not feasible, administrators should restrict the ability of users to import external or untrusted model checkpoints.