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

Designing effective defense against adversarial attacks is a crucial topic asdeep neural networks have been proliferated rapidly in many security-criticaldomains such as malware detection and self-driving cars. Conventional defensemethods, although shown to be promising, are largely limited by theirsingle-source single-cost nature: The robustness promotion tends to plateauwhen the defenses are made increasingly stronger while the cost tends toamplify. In this paper, we study principles of designing multi-source andmulti-cost schemes where defense performance is boosted from multiple defendingcomponents. Based on this motivation, we propose a multi-source and multi-costdefense scheme, Adversarially Trained Model Switching (AdvMS), that inheritsadvantages from two leading schemes: adversarial training and random modelswitching. We show that the multi-source nature of AdvMS mitigates theperformance plateauing issue and the multi-cost nature enables improvingrobustness at a flexible and adjustable combination of costs over differentfactors which can better suit specific restrictions and needs in practice.

AdvMS: A Multi-source Multi-cost Defense Against Adversarial Attacks