dc.contributor.author Wu, Kaiwen dc.date.accessioned 2020-09-21 20:27:51 (GMT) dc.date.available 2020-09-21 20:27:51 (GMT) dc.date.issued 2020-09-21 dc.date.submitted 2020-09-16 dc.identifier.uri http://hdl.handle.net/10012/16345 dc.description.abstract Deep models, while being extremely flexible and accurate, are surprisingly vulnerable to small, imperceptible'' perturbations known as adversarial attacks. While the majority of existing attacks focus on measuring perturbations under the $\ell_p$ metric, Wasserstein distance, which takes geometry in pixel space into account, has long been known to be a suitable metric for measuring image quality and has recently risen as a compelling alternative to the $\ell_p$ metric in adversarial attacks. However, constructing an effective attack under the Wasserstein metric is computationally much more challenging and calls for better optimization algorithms. We address this gap in two ways: (a) we develop an exact yet efficient projection operator to enable a stronger projected gradient attack; (b) we show that the Frank-Wolfe method equipped with a suitable linear minimization oracle works extremely fast under Wasserstein constraints. Our algorithms not only converge faster but also generate much stronger attacks. For instance, we decrease the accuracy of a residual network on CIFAR-10 to 3.4% within a Wasserstein perturbation ball of radius 0.005, in contrast to 65.6% using the previous Wasserstein attack based on an approximate projection operator. Furthermore, employing our stronger attacks in adversarial training significantly improves the robustness of adversarially trained models. Our algorithms are applicable to general Wasserstein constrained optimization problems in other domains beyond adversarial robustness. en dc.language.iso en en dc.publisher University of Waterloo en dc.relation.uri https://github.com/watml/fast-wasserstein-adversarial en dc.subject Wasserstein distance en dc.subject adversarial robustness en dc.subject optimization en dc.title Wasserstein Adversarial Robustness en dc.type Master Thesis en dc.pending false uws-etd.degree.department David R. Cheriton School of Computer Science en uws-etd.degree.discipline Computer Science en uws-etd.degree.grantor University of Waterloo en uws-etd.degree Master of Mathematics en uws.contributor.advisor Yu, Yaoliang uws.contributor.affiliation1 Faculty of Mathematics en uws.published.city Waterloo en uws.published.country Canada en uws.published.province Ontario en uws.typeOfResource Text en uws.peerReviewStatus Unreviewed en uws.scholarLevel Graduate en
﻿

### This item appears in the following Collection(s)

UWSpace

University of Waterloo Library
200 University Avenue West