Generalization Properties of Adversarial Training for $\ell_0$-Bounded Adversarial Attacks

We have widely observed that neural networks are vulnerable to small additive perturbations to the input causing misclassification. In this paper, we focus on the $\ell_0$-bounded adversarial attacks, and aim to theoretically characterize the performance of adversarial training for an important class of truncated classifiers. Such classifiers are shown to have strong performance empirically, as well as theoretically in the Gaussian mixture model, in the $\ell_0$-adversarial setting. The main contribution of this paper is to prove a novel generalization bound …

adversarial adversarial attacks aim attacks class cs.cr cs.lg focus important input networks neural networks performance training truncated vulnerable