Rethinking Model Ensemble in Transfer-based Adversarial Attacks

TL;DR

This paper analyzes why ensemble methods improve transfer-based adversarial attacks and introduces a new attack strategy, CWA, that enhances transferability by targeting model weaknesses related to loss landscape flatness and local optima proximity.

Contribution

It provides a theoretical and empirical analysis of ensemble weaknesses and proposes the Common Weakness Attack (CWA) to generate more transferable adversarial examples.

Findings

CWA improves transferability on image classification tasks.

CWA enhances attack success on object detection models.

Effective against adversarially trained models and real-world systems like Google's Bard.

Abstract

It is widely recognized that deep learning models lack robustness to adversarial examples. An intriguing property of adversarial examples is that they can transfer across different models, which enables black-box attacks without any knowledge of the victim model. An effective strategy to improve the transferability is attacking an ensemble of models. However, previous works simply average the outputs of different models, lacking an in-depth analysis on how and why model ensemble methods can strongly improve the transferability. In this paper, we rethink the ensemble in adversarial attacks and define the common weakness of model ensemble with two properties: 1) the flatness of loss landscape; and 2) the closeness to the local optimum of each model. We empirically and theoretically show that both properties are strongly correlated with the transferability and propose a Common Weakness Attack (CWA) to generate more transferable adversarial examples by promoting these two properties. Experimental results on both image classification and object detection tasks validate the effectiveness of our approach to improving the adversarial transferability, especially when attacking adversarially trained models. We also successfully apply our method to attack a black-box large vision-language model -- Google's Bard, showing the practical effectiveness. Code is available at \url{https://github.com/huanranchen/AdversarialAttacks}.