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/

Zendy Plus

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

Zendy Tools

Zendy Open

Adversarial examples are inputs with imperceptible perturbations that easilymisleading deep neural networks(DNNs). Recently, adversarial patch, with noiseconfined to a small and localized patch, has emerged for its easy feasibilityin real-world scenarios. However, existing strategies failed to generateadversarial patches with strong generalization ability. In other words, theadversarial patches were input-specific and failed to attack images from allclasses, especially unseen ones during training. To address the problem, thispaper proposes a bias-based framework to generate class-agnostic universaladversarial patches with strong generalization ability, which exploits both theperceptual and semantic bias of models. Regarding the perceptual bias, sinceDNNs are strongly biased towards textures, we exploit the hard examples whichconvey strong model uncertainties and extract a textural patch prior from themby adopting the style similarities. The patch prior is more close to decisionboundaries and would promote attacks. To further alleviate the heavy dependencyon large amounts of data in training universal attacks, we further exploit thesemantic bias. As the class-wise preference, prototypes are introduced andpursued by maximizing the multi-class margin to help universal training. TakingAutomaticCheck-out (ACO) as the typical scenario, extensive experimentsincluding white-box and black-box settings in both digital-world(RPC, thelargest ACO related dataset) and physical-world scenario(Taobao and JD, theworld' s largest online shopping platforms) are conducted. Experimental resultsdemonstrate that our proposed framework outperforms state-of-the-artadversarial patch attack methods.

Bias-based Universal Adversarial Patch Attack for Automatic Check-out