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 security threat to deep neural networks (DNN) is backdoor contamination, inwhich an adversary poisons the training data of a target model to inject aTrojan so that images carrying a specific trigger will always be classifiedinto a specific label. Prior research on this problem assumes the dominance ofthe trigger in an image's representation, which causes any image with thetrigger to be recognized as a member in the target class. Such a trigger alsoexhibits unique features in the representation space and can therefore beeasily separated from legitimate images. Our research, however, shows thatsimple target contamination can cause the representation of an attack image tobe less distinguishable from that of legitimate ones, thereby evading existingdefenses against the backdoor infection.  In our research, we show that such a contamination attack actually subtlychanges the representation distribution for the target class, which can becaptured by a statistic analysis. More specifically, we leverage an EMalgorithm to decompose an image into its identity part (e.g., person, trafficsign) and variation part within a class (e.g., lighting, poses). Then weanalyze the distribution in each class, identifying those more likely to becharacterized by a mixture model resulted from adding attack samples to thelegitimate image pool. Our research shows that this new technique effectivelydetects data contamination attacks, including the new one we propose, and isalso robust against the evasion attempts made by a knowledgeable adversary.

Demon in the Variant: Statistical Analysis of DNNs for Robust Backdoor Contamination Detection