Learning with a Strong Adversary
This addresses the issue of adversarial vulnerability in neural networks, which is critical for security-sensitive applications, but the method appears incremental as it builds on existing adversarial training approaches.
The paper tackles the problem of neural network robustness to adversarial perturbations by proposing a method that uses adversarial examples as an intermediate step, resulting in greatly improved robustness for classification models.
The robustness of neural networks to intended perturbations has recently attracted significant attention. In this paper, we propose a new method, \emph{learning with a strong adversary}, that learns robust classifiers from supervised data. The proposed method takes finding adversarial examples as an intermediate step. A new and simple way of finding adversarial examples is presented and experimentally shown to be efficient. Experimental results demonstrate that resulting learning method greatly improves the robustness of the classification models produced.