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 patch attacks can fool the face recognition (FR) models via smallpatches. However, previous adversarial patch attacks often result in unnaturalpatterns that are easily noticeable. Generating transferable and stealthyadversarial patches that can efficiently deceive the black-box FR models whilehaving good camouflage is challenging because of the huge stylistic differencebetween the source and target images. To generate transferable,natural-looking, and stealthy adversarial patches, we propose an innovativetwo-stage attack called Adv-Inpainting, which extracts style features andidentity features from the attacker and target faces, respectively and thenfills the patches with misleading and inconspicuous content guided by attentionmaps. In the first stage, we extract multi-scale style embeddings by apyramid-like network and identity embeddings by a pretrained FR model andpropose a novel Attention-guided Adaptive Instance Normalization layer (AAIN)to merge them via background-patch cross-attention maps. The proposed layer canadaptively fuse identity and style embeddings by fully exploiting prioritycontextual information. In the second stage, we design an Adversarial PatchRefinement Network (APR-Net) with a novel boundary variance loss, a spatialdiscounted reconstruction loss, and a perceptual loss to boost the stealthinessfurther. Experiments demonstrate that our attack can generate adversarialpatches with improved visual quality, better stealthiness, and strongertransferability than state-of-the-art adversarial patch attacks and semanticattacks.

Generating Transferable and Stealthy Adversarial Patch via Attention-guided Adversarial Inpainting