The Adversarial Robustness Toolbox (ART) through version 1.20.1 contains a command-line argument injection vulnerability in its Kubeflow component (robustness_evaluation_fgsm_pytorch.py). The script uses the unsafe eval() function to parse string values provided via the --clip_values and --input_shape command-line arguments. This unsafe use allows an attacker who can control these arguments, for example through pipeline configuration or automated scripts, to inject and execute arbitrary Python code on the system running the ART evaluation. The vulnerability is rated critical with a CVSS score of 9.8, reflecting network attack vector, low complexity, no privileges or user interaction required, and high impact on confidentiality, integrity, and availability. The vendor advisory from Red Hat confirms the vulnerability and recommends restricting access controls to prevent unauthorized users from controlling these arguments. No official patch or fix is currently confirmed.

Successful exploitation of this vulnerability allows an attacker to execute arbitrary Python code remotely on the system running the ART evaluation scripts. This can lead to full compromise of the affected system, including unauthorized access, data integrity loss, and denial of service. The vulnerability affects confidentiality, integrity, and availability at a high level. There are no known exploits in the wild at this time. The vulnerability is critical due to its ease of exploitation and severe impact.

No official patch or fix is currently confirmed for this vulnerability. The vendor advisory recommends mitigating the risk by ensuring that only trusted and authorized users have permissions to create, modify, or deploy Kubeflow pipelines and configurations within Red Hat OpenShift AI environments where ART is used. Restricting access to the environment where the ART evaluation scripts run prevents attackers from controlling the vulnerable command-line arguments. These access controls should be maintained persistently, including after service restarts or reloads.