Detecting Adversarial Examples Is (Nearly) As Hard As Classifying Them

摘要

Making classifiers robust to adversarial examples is challenging. Thus, many works tackle the seemingly easier task of