Improving Resistance to Adversarial Deformations by Regularizing Gradients

Pengfei Xia, Bin Li

Improving the resistance of deep neural networks against adversarial attacks is important for deploying models to realistic applications. However, most defense methods are designed…