Foolbox

An open-source adversarial robustness evaluation library compatible with major deep learning frameworks including PyTorch and TensorFlow. It provides a range of adversarial attack algorithms and is primarily used for security testing of image classification models to assess their resilience against adversarial inputs.