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 vulnerability of deep neural networks to adversarial examples hasmotivated an increasing number of defense strategies for promoting modelrobustness. However, the progress is usually hampered by insufficientrobustness evaluations. As the de facto standard to evaluate adversarialrobustness, adversarial attacks typically solve an optimization problem ofcrafting adversarial examples with an iterative process. In this work, wepropose a Model-Agnostic Meta-Attack (MAMA) approach to discover strongerattack algorithms automatically. Our method learns the optimizer in adversarialattacks parameterized by a recurrent neural network, which is trained over aclass of data samples and defenses to produce effective update directionsduring adversarial example generation. Furthermore, we develop a model-agnostictraining algorithm to improve the generalization ability of the learnedoptimizer when attacking unseen defenses. Our approach can be flexiblyincorporated with various attacks and consistently improves the performancewith little extra computational cost. Extensive experiments demonstrate theeffectiveness of the learned attacks by MAMA compared to the state-of-the-artattacks on different defenses, leading to a more reliable evaluation ofadversarial robustness.

Model-Agnostic Meta-Attack: Towards Reliable Evaluation of Adversarial Robustness