English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

In this technical report, we evaluate the adversarial robustness of a veryrecent method called "Geometry-aware Instance-reweighted AdversarialTraining"[7]. GAIRAT reports state-of-the-art results on defenses toadversarial attacks on the CIFAR-10 dataset. In fact, we find that a networktrained with this method, while showing an improvement over regular adversarialtraining (AT), is biasing the model towards certain samples by re-scaling theloss. Indeed, this leads the model to be susceptible to attacks that scale thelogits. The original model shows an accuracy of 59% under AutoAttack - whentrained with additional data with pseudo-labels. We provide an analysis thatshows the opposite. In particular, we craft a PGD attack multiplying the logitsby a positive scalar that decreases the GAIRAT accuracy from from 55% to 44%,when trained solely on CIFAR-10. In this report, we rigorously evaluate themodel and provide insights into the reasons behind the vulnerability of GAIRATto this adversarial attack. The code to reproduce our evaluation is madeavailable at https://github.com/giuxhub/GAIRAT-LSA

Evaluating the Robustness of Geometry-Aware Instance-Reweighted Adversarial Training