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

The security of Deep Learning classifiers is a critical field of studybecause of the existence of adversarial attacks. Such attacks usually rely onthe principle of transferability, where an adversarial example crafted on asurrogate classifier tends to mislead the target classifier trained on the samedataset even if both classifiers have quite different architecture. Ensemblemethods against adversarial attacks demonstrate that an adversarial example isless likely to mislead multiple classifiers in an ensemble having diversedecision boundaries. However, recent ensemble methods have either been shown tobe vulnerable to stronger adversaries or shown to lack an end-to-endevaluation. This paper attempts to develop a new ensemble methodology thatconstructs multiple diverse classifiers using a Pairwise Adversarially RobustLoss (PARL) function during the training procedure. PARL utilizes gradients ofeach layer with respect to input in every classifier within the ensemblesimultaneously. The proposed training procedure enables PARL to achieve higherrobustness against black-box transfer attacks compared to previous ensemblemethods without adversely affecting the accuracy of clean examples. We alsoevaluate the robustness in the presence of white-box attacks, where adversarialexamples are crafted using parameters of the target classifier. We presentextensive experiments using standard image classification datasets likeCIFAR-10 and CIFAR-100 trained using standard ResNet20 classifier againststate-of-the-art adversarial attacks to demonstrate the robustness of theproposed ensemble methodology.

PARL: Enhancing Diversity of Ensemble Networks to Resist Adversarial Attacks via Pairwise Adversarially Robust Loss Function