Enhance Adversarial Robustness via Geodesic Distance
- 9 February 2024
- journal article
- research article
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Artificial Intelligence
- Vol. 5 (8) , 4202-4216
Abstract
Adversarial training is an effective method to improve the model’s adversarial robustness. To realize a considerable tradeoff between clean accuracy and adversarial robustness, surrogate loss minimization can be utilized to regularize for robustness. This study advances beyond previous research efforts by taking a further step. It is informed by two theoretical perspectives. First, adversarial examples are inevitable within a unit sphere surrounding clean data. Second, geodesics can characterize the shortest distance between two points in Riemannian geometry. Accordingly, this study employs geodesic distance as a regularized surrogate loss item to capture the minimal divergence between the distribution of natural examples and the distribution of adversarial examples. This approach yields a more compact upper bound of risk errors than previous studies, which is beneficial for improving adversarial robustness. Based on the theoretical insight, this study proposes a metric of Geodesic Loss and a framework of Geodesic Adversarial Training to boost the adversarial robustness of neural networks. The empirical study on the diverse datasets demonstrates the considerable performance of the proposed method against diverse attacks, including white-box attacks, black-box corruptions, adaptive attacks, and auto attacks. Our code is available at: https://github.com/momo1986/GeodesicAdversarialTraining .Keywords
Funding Information
- National Natural Science Foundation of China (62133011)
This publication has 39 references indexed in Scilit:
- Skilful precipitation nowcasting using deep generative models of radarNature, 2021
- Boosting Decision-Based Black-Box Adversarial Attacks with Random Sign Flip2020
- Improved protein structure prediction using potentials from deep learningNature, 2020
- Recent Advances in Big Data and Deep LearningProceedings of the International Neural Networks Society, 2020
- Clinically applicable deep learning framework for organs at risk delineation in CT imagesNature Machine Intelligence, 2019
- Riemannian Walk for Incremental Learning: Understanding Forgetting and Intransigence2018
- Analysis of classifiers’ robustness to adversarial perturbationsMachine Learning, 2017
- Optimal Transport2014
- Metrics for 3D Rotations: Comparison and AnalysisJournal of Mathematical Imaging and Vision, 2009
- The Bayesian Approach to Local Optimization1989