An Information-Theoretic Explanation for the Adversarial Fragility of AI Classifiers | IEEE Conference Publication | IEEE Xplore