Robust Principles: Architectural Design Principles for Adversarially Robust CNNs

ShengYun Peng, Weilin Xu, Cory Cornelius, Matthew Hull, Kevin Li +4 more
8/30/2023
cs.CV

Official code: poloclub/robust-principles (24 stars, MIT, last commit 2024-01-13).

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.

View on arXivView PDF

Code Implementations(1)

poloclub/robust-principlesOfficial100%
MIT

Robust Principles: Architectural Design Principles for Adversarially Robust CNNs

244Python, MakefileAug 26, 20232 years agoMIT
adversarial-attacksadversarial-machine-learningarchitecturerobustness

Similar Papers

RobArch: Designing Robust Architectures against Adversarial Attacks

87% match1 repo42023

NERO-Net: A Neuroevolutionary Approach for the Design of Adversarially Robust CNNs

85% match2026

Improving Generalization of Adversarial Training via Robust Critical Fine-Tuning

81% match2 repos5082023

RBFormer: Improve Adversarial Robustness of Transformer by Robust Bias

80% match2023

Neural Architecture Design and Robustness: A Dataset

80% match1 repo162023

What Makes VLMs Robust? Towards Reconciling Robustness and Accuracy in Vision-Language Models

80% match2026

Cite this paper

@article{peng2023robust,
  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},
  year   = {2023},
  eprint = {2308.16258},
  archivePrefix = {arXiv},
  url    = {http://arxiv.org/abs/2308.16258v2}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/2308.16258.svg)](https://codeofpaper.com/paper/2308.16258)

Discussion