Robust Principles: Architectural Design Principles for Adversarially Robust CNNs
Abstract
Our research aims to unify existing works' diverging opinions on how architectural components affect the adversarial robustness of CNNs. To accomplish our goal, we synthesize a suite of three generalizable robust architectural design principles: (a) optimal range for depth and width configurations, (b) preferring convolutional over patchify stem stage, and (c) robust residual block design through adopting squeeze and excitation blocks and non-parametric smooth activation functions. Through extensive experiments across a wide spectrum of dataset scales, adversarial training methods, model parameters, and network design spaces, our principles consistently and markedly improve AutoAttack accuracy: 1-3 percentage points (pp) on CIFAR-10 and CIFAR-100, and 4-9 pp on ImageNet. The code is publicly available at https://github.com/poloclub/robust-principles.
Cite
@article{arxiv.2308.16258,
title = {Robust Principles: Architectural Design Principles for Adversarially Robust CNNs},
author = {ShengYun Peng and Weilin Xu and Cory Cornelius and Matthew Hull and Kevin Li and Rahul Duggal and Mansi Phute and Jason Martin and Duen Horng Chau},
journal= {arXiv preprint arXiv:2308.16258},
year = {2023}
}
Comments
Published at BMVC'23