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

A significant threat to the recent, wide deployment of machine learning-basedsystems, including deep neural networks (DNNs), is adversarial learningattacks. We analyze possible test-time evasion-attack mechanisms and show that,in some important cases, when the image has been attacked, correctlyclassifying it has no utility: i) when the image to be attacked is (evenarbitrarily) selected from the attacker's cache; ii) when the sole recipient ofthe classifier's decision is the attacker. Moreover, in some applicationdomains and scenarios it is highly actionable to detect the attack irrespectiveof correctly classifying in the face of it (with classification still performedif no attack is detected). We hypothesize that, even if human-imperceptible,adversarial perturbations are machine-detectable. We propose a purelyunsupervised anomaly detector (AD) that, unlike previous works: i) models thejoint density of a deep layer using highly suitable null hypothesis densitymodels (matched in particular to the non- negative support for RELU layers);ii) exploits multiple DNN layers; iii) leverages a "source" and "destination"class concept, source class uncertainty, the class confusion matrix, and DNNweight information in constructing a novel decision statistic grounded in theKullback-Leibler divergence. Tested on MNIST and CIFAR-10 image databases underthree prominent attack strategies, our approach outperforms previous detectionmethods, achieving strong ROC AUC detection accuracy on two attacks and betteraccuracy than recently reported for a variety of methods on the strongest (CW)attack. We also evaluate a fully white box attack on our system. Finally, weevaluate other important performance measures, such as classification accuracy,versus detection rate and attack strength.

When Not to Classify: Anomaly Detection of Attacks (ADA) on DNN Classifiers at Test Time