The fact that deep neural networks are susceptible to crafted perturbations\nseverely impacts the use of deep learning in certain domains of application.\nAmong many developed defense models against such attacks, adversarial training\nemerges as the most successful method that consistently resists a wide range of\nattacks. In this work, based on an observation from a previous study that the\nrepresentations of a clean data example and its adversarial examples become\nmore divergent in higher layers of a deep neural net, we propose the Adversary\nDivergence Reduction Network which enforces local/global compactness and the\nclustering assumption over an intermediate layer of a deep neural network. We\nconduct comprehensive experiments to understand the isolating behavior of each\ncomponent (i.e., local/global compactness and the clustering assumption) and\ncompare our proposed model with state-of-the-art adversarial training methods.\nThe experimental results demonstrate that augmenting adversarial training with\nour proposed components can further improve the robustness of the network,\nleading to higher unperturbed and adversarial predictive performances.\n