Mesh Attack

Mesh attacks are adversarial attacks targeting 3D mesh-based machine learning models, aiming to subtly alter mesh geometry to cause misclassification or incorrect reconstruction. Current research focuses on developing effective attack methods, often leveraging spectral analysis or surface-based perturbations, and evaluating their efficacy against various deep learning architectures like PointNet and DGCNNs, as well as autoencoders. These studies are crucial for assessing the robustness of 3D shape recognition and processing systems, ultimately improving the security and reliability of applications ranging from facial recognition to computer-aided design.