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 Tools annual

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 Tools monthly

Global: Zendy Plus monthly

Presents the access of premium content as premium feature

Premium Content

Global: Zendy Plus annual

Global: Zendy Free Plan

The field of defense strategies against adversarial attacks has significantlygrown over the last years, but progress is hampered as the evaluation ofadversarial defenses is often insufficient and thus gives a wrong impression ofrobustness. Many promising defenses could be broken later on, making itdifficult to identify the state-of-the-art. Frequent pitfalls in the evaluationare improper tuning of hyperparameters of the attacks, gradient obfuscation ormasking. In this paper we first propose two extensions of the PGD-attackovercoming failures due to suboptimal step size and problems of the objectivefunction. We then combine our novel attacks with two complementary existingones to form a parameter-free, computationally affordable and user-independentensemble of attacks to test adversarial robustness. We apply our ensemble toover 50 models from papers published at recent top machine learning andcomputer vision venues. In all except one of the cases we achieve lower robusttest accuracy than reported in these papers, often by more than $10\%$,identifying several broken defenses.

Reliable evaluation of adversarial robustness with an ensemble of diverse parameter-free attacks