On the Adversarial Transferability of ConvMixer Models

TL;DR

This paper investigates how adversarial examples transfer between ConvMixer models and other neural networks, revealing that ConvMixer models are vulnerable to transfer attacks, which has implications for their robustness in image classification.

Contribution

First study to analyze adversarial transferability involving ConvMixer models using a benchmark attack method, highlighting their vulnerability.

Findings

ConvMixer models are weak to adversarial transferability

AutoAttack effectively evaluates model robustness

Transferability poses security concerns for ConvMixer networks

Abstract

Deep neural networks (DNNs) are well known to be vulnerable to adversarial examples (AEs). In addition, AEs have adversarial transferability, which means AEs generated for a source model can fool another black-box model (target model) with a non-trivial probability. In this paper, we investigate the property of adversarial transferability between models including ConvMixer, which is an isotropic network, for the first time. To objectively verify the property of transferability, the robustness of models is evaluated by using a benchmark attack method called AutoAttack. In an image classification experiment, ConvMixer is confirmed to be weak to adversarial transferability.