The Adversarial Robustness Toolbox (ART) through version 1.20.1 contains a remote code execution vulnerability in its Kubeflow component. The vulnerability is due to the use of the unsafe eval() function in the robustness evaluation function for PyTorch models, which dynamically evaluates user-supplied strings for LossFn and Optimizer parameters without any sanitization or security restrictions. This allows an attacker to execute arbitrary Python code on the system running the ART evaluation by providing specially crafted input strings.

Successful exploitation of this vulnerability allows an attacker to execute arbitrary Python code remotely on the affected system, leading to complete compromise of the system running the ART evaluation. This can result in unauthorized access, data manipulation, or disruption of services.

Patch status is not yet confirmed — check the vendor advisory for current remediation guidance. Until an official fix is available, users should avoid running the vulnerable ART Kubeflow component with untrusted input for LossFn and Optimizer parameters. Implementing input validation or sandboxing the evaluation environment may reduce risk but are not guaranteed mitigations.