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

Adversarial detection aims to determine whether a given sample is anadversarial one based on the discrepancy between natural and adversarialdistributions. Unfortunately, estimating or comparing two data distributions isextremely difficult, especially in high-dimension spaces. Recently, thegradient of log probability density (a.k.a., score) w.r.t. the sample is usedas an alternative statistic to compute. However, we find that the score issensitive in identifying adversarial samples due to insufficient informationwith one sample only. In this paper, we propose a new statistic called expectedperturbation score (EPS), which is essentially the expected score of a sampleafter various perturbations. Specifically, to obtain adequate informationregarding one sample, we perturb it by adding various noises to capture itsmulti-view observations. We theoretically prove that EPS is a proper statisticto compute the discrepancy between two samples under mild conditions. Inpractice, we can use a pre-trained diffusion model to estimate EPS for eachsample. Last, we propose an EPS-based adversarial detection (EPS-AD) method, inwhich we develop EPS-based maximum mean discrepancy (MMD) as a metric tomeasure the discrepancy between the test sample and natural samples. We alsoprove that the EPS-based MMD between natural and adversarial samples is largerthan that among natural samples. Extensive experiments show the superioradversarial detection performance of our EPS-AD.

Detecting Adversarial Data by Probing Multiple Perturbations Using Expected Perturbation Score