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

Deep-learning based classification algorithms have been shown to besusceptible to adversarial attacks: minor changes to the input of classifierscan dramatically change their outputs, while being imperceptible to humans. Inthis paper, we present a simple hypothesis about a feature compression propertyof artificial intelligence (AI) classifiers and present theoretical argumentsto show that this hypothesis successfully accounts for the observed fragilityof AI classifiers to small adversarial perturbations. Drawing on ideas frominformation and coding theory, we propose a general class of defenses fordetecting classifier errors caused by abnormally small input perturbations. Wefurther show theoretical guarantees for the performance of this detectionmethod. We present experimental results with (a) a voice recognition system,and (b) a digit recognition system using the MNIST database, to demonstrate theeffectiveness of the proposed defense methods. The ideas in this paper aremotivated by a simple analogy between AI classifiers and the standard Shannonmodel of a communication system.

Trust but Verify: An Information-Theoretic Explanation for the  Adversarial Fragility of Machine Learning Systems, and a General Defense  against Adversarial Attacks