Impact of Attention on Adversarial Robustness of Image Classification Models

TL;DR

This study compares the adversarial robustness of attention-based and non-attention image classification models across multiple datasets and attack types, revealing dataset-dependent robustness improvements with attention mechanisms.

Contribution

It provides a comprehensive comparative analysis of attention versus non-attention models' robustness against adversarial attacks on several datasets.

Findings

Attention models show improved robustness on datasets with fewer classes.

Robustness of attention models varies depending on dataset complexity.

Dataset characteristics influence the effectiveness of attention mechanisms against attacks.

Abstract

Adversarial attacks against deep learning models have gained significant attention and recent works have proposed explanations for the existence of adversarial examples and techniques to defend the models against these attacks. Attention in computer vision has been used to incorporate focused learning of important features and has led to improved accuracy. Recently, models with attention mechanisms have been proposed to enhance adversarial robustness. Following this context, this work aims at a general understanding of the impact of attention on adversarial robustness. This work presents a comparative study of adversarial robustness of non-attention and attention based image classification models trained on CIFAR-10, CIFAR-100 and Fashion MNIST datasets under the popular white box and black box attacks. The experimental results show that the robustness of attention based models may be dependent on the datasets used i.e. the number of classes involved in the classification. In contrast to the datasets with less number of classes, attention based models are observed to show better robustness towards classification.