TY - GEN
T1 - Escaping Adversarial Attacks with Egyptian Mirrors
AU - Saukh, Olga
N1 - Publisher Copyright:
© 2023 Owner/Author.
PY - 2024/3/12
Y1 - 2024/3/12
N2 - Adversarial robustness received significant attention over the past years, due to its critical practical role. Complementary to the existing literature on adversarial training, we explore weight-space ensembles of independently trained models. We propose a defense against adversarial examples which takes advantage of the latest empirical findings on linear mode connectivity of overparameterized models modulo permutation invariance. Egyptian Mirrors defense escapes adversarial attacks by moving along linear paths between pair-wise aligned functionally diverse models, while frequently and arbitrary changing ensembling direction. We evaluate the proposed defense using adversarial examples generated by FGSM and PGD attacks and show improvements up to 8% and 33% test accuracy on 2-layer MLP and VGG11 architectures trained on GTSRB and CIFAR10 datasets respectively.
AB - Adversarial robustness received significant attention over the past years, due to its critical practical role. Complementary to the existing literature on adversarial training, we explore weight-space ensembles of independently trained models. We propose a defense against adversarial examples which takes advantage of the latest empirical findings on linear mode connectivity of overparameterized models modulo permutation invariance. Egyptian Mirrors defense escapes adversarial attacks by moving along linear paths between pair-wise aligned functionally diverse models, while frequently and arbitrary changing ensembling direction. We evaluate the proposed defense using adversarial examples generated by FGSM and PGD attacks and show improvements up to 8% and 33% test accuracy on 2-layer MLP and VGG11 architectures trained on GTSRB and CIFAR10 datasets respectively.
UR - https://www.scopus.com/pages/publications/85188815307
U2 - 10.1145/3615593.3615724
DO - 10.1145/3615593.3615724
M3 - Conference paper
AN - SCOPUS:85188815307
T3 - ACM MobiCom 2023 - Proceedings of the 2023 2nd ACM Workshop on Data Privacy and Federated Learning Technologies for Mobile Edge Network, FedEdge 2023
SP - 131
EP - 136
BT - ACM MobiCom 2023 - Proceedings of the 2023 2nd ACM Workshop on Data Privacy and Federated Learning Technologies for Mobile Edge Network, FedEdge 2023
PB - Association for Computing Machinery (ACM)
T2 - 2nd ACM Workshop on Data Privacy and Federated Learning Technologies for Mobile Edge Network, FedEdge 2023, Collocated with ACM MobiCom 2023
Y2 - 2 October 2023 through 2 October 2023
ER -