Symmetry Defense Against CNN Adversarial Perturbation Attacks

Blerta Lindqvist*

*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceedingConference article in proceedingsScientificpeer-review

1 Citation (Scopus)

Abstract

This paper uses symmetry to make Convolutional Neural Network classifiers (CNNs) robust against adversarial perturbation attacks. Such attacks add perturbation to original images to generate adversarial images that fool classifiers such as road sign classifiers of autonomous vehicles. Although symmetry is a pervasive aspect of the natural world, CNNs are unable to handle symmetry well. For example, a CNN can classify an image differently from its mirror image. For an adversarial image that misclassifies with a wrong label lw, CNN inability to handle symmetry means that a symmetric adversarial image can classify differently from the wrong label lw. Further than that, we find that the classification of a symmetric adversarial image reverts to the correct label. To classify an image when adversaries are unaware of the defense, we apply symmetry to the image and use the classification label of the symmetric image. To classify an image when adversaries are aware of the defense, we use mirror symmetry and pixel inversion symmetry to form a symmetry group. We apply all the group symmetries to the image and decide on the output label based on the agreement of any two of the classification labels of the symmetry images. Adaptive attacks fail because they need to rely on loss functions that use conflicting CNN output values for symmetric images. Without attack knowledge, the proposed symmetry defense succeeds against both gradient-based and random-search attacks, with up to near-default accuracies for ImageNet. The defense even improves the classification accuracy of original images.

Original languageEnglish
Title of host publicationInformation Security - 26th International Conference, ISC 2023, Proceedings
EditorsElias Athanasopoulos, Bart Mennink
PublisherSpringer
Pages142-160
Number of pages19
ISBN (Print)9783031491863
DOIs
Publication statusPublished - 2023
MoE publication typeA4 Conference publication
EventInformation Security Conference - Groningen, Netherlands
Duration: 15 Nov 202317 Nov 2023
Conference number: 26

Publication series

NameLecture Notes in Computer Science
Volume14411
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

ConferenceInformation Security Conference
Abbreviated titleISC
Country/TerritoryNetherlands
CityGroningen
Period15/11/202317/11/2023

Keywords

  • Adversarial perturbation defense
  • CNN adversarial robustness
  • Symmetry

Fingerprint

Dive into the research topics of 'Symmetry Defense Against CNN Adversarial Perturbation Attacks'. Together they form a unique fingerprint.

Cite this