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

Recently, physical adversarial attacks have been presented to evadeDNNs-based object detectors. To ensure the security, many scenarios aresimultaneously deployed with visible sensors and infrared sensors, leading tothe failures of these single-modal physical attacks. To show the potentialrisks under such scenes, we propose a unified adversarial patch to performcross-modal physical attacks, i.e., fooling visible and infrared objectdetectors at the same time via a single patch. Considering different imagingmechanisms of visible and infrared sensors, our work focuses on modeling theshapes of adversarial patches, which can be captured in different modalitieswhen they change. To this end, we design a novel boundary-limited shapeoptimization to achieve the compact and smooth shapes, and thus they can beeasily implemented in the physical world. In addition, to balance the foolingdegree between visible detector and infrared detector during the optimizationprocess, we propose a score-aware iterative evaluation, which can guide theadversarial patch to iteratively reduce the predicted scores of the multi-modalsensors. We finally test our method against the one-stage detector: YOLOv3 andthe two-stage detector: Faster RCNN. Results show that our unified patchachieves an Attack Success Rate (ASR) of 73.33% and 69.17%, respectively. Moreimportantly, we verify the effective attacks in the physical world when visibleand infrared sensors shoot the objects under various settings like differentangles, distances, postures, and scenes.

Unified Adversarial Patch for Cross-modal Attacks in the Physical World