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 attacks pose serious challenges for deep neural network(DNN)-based analysis of various input signals. In the case of three-dimensionalpoint clouds, methods have been developed to identify points that play a keyrole in network decision, and these become crucial in generating existingadversarial attacks. For example, a saliency map approach is a popular methodfor identifying adversarial drop points, whose removal would significantlyimpact the network decision. This paper seeks to enhance the understanding ofthree-dimensional adversarial attacks by exploring which point cloud featuresare most important for predicting adversarial points. Specifically, Fourteenkey point cloud features such as edge intensity and distance from the centroidare defined, and multiple linear regression is employed to assess theirpredictive power for adversarial points. Based on critical feature selectioninsights, a new attack method has been developed to evaluate whether theselected features can generate an attack successfully. Unlike traditionalattack methods that rely on model-specific vulnerabilities, this approachfocuses on the intrinsic characteristics of the point clouds themselves. It isdemonstrated that these features can predict adversarial points across fourdifferent DNN architectures, Point Network (PointNet), PointNet++, DynamicGraph Convolutional Neural Networks (DGCNN), and Point Convolutional Network(PointConv) outperforming random guessing and achieving results comparable tosaliency map-based attacks. This study has important engineering applications,such as enhancing the security and robustness of three-dimensional pointcloud-based systems in fields like robotics and autonomous driving.

Understanding Key Point Cloud Features for Development Three-dimensional Adversarial Attacks