CIFAR-10 and NIPS-W

The dataset used in the paper is CIFAR-10 and NIPS-W, which are benchmark datasets for image classification and adversarial attacks.

BibTex: