Random Walks for Adversarial Meshes
Event Type
Technical Paper
Interest Areas
Research & Education
Presentation Types
Registration Categories
Full Conference Supporter
Full Conference
Virtual Conference Supporter
Virtual Conference
Exhibitor Additional Full Conference
Exhibitor Full Conference
DescriptionWe propose a novel, unified, and general adversarial attack, which leads to misclassification of numerous SOTA mesh classification neural networks. The attack is black-box, having access only to the network’s predictions but not to its architecture or gradients. The key idea is to learn to imitate a given classification network.