Defense-GAN: Protecting Classifiers Against Adversarial Attacks Using Generative Models (published in ICLR2018) - View it on GitHub