adversarial-robustness-toolbox

The Adversarial Robustness Toolbox (ART) is a Python library designed to enhance the security of machine learning models by providing tools to defend against various adversarial threats, including Evasion, Poisoning, Extraction, and Inference attacks. It supports a wide range of popular machine learning frameworks and data types, enabling researchers and developers to evaluate and improve the robustness of their applications across multiple tasks such as classification, object detection, and speech recognition. Notable features include compatibility with numerous ML frameworks and the capability to handle diverse data formats, making it versatile for various machine learning security applications.