You can trick a 3D AI by changing the 'holes' and connections in an object while keeping its shape looking perfectly normal to a human.
April 14, 2026
Original Paper
Topo-ADV: Generating Topology-Driven Imperceptible Adversarial Point Clouds
arXiv · 2604.09879
The Takeaway
This method manipulates the topological (homological) structure of point clouds to create imperceptible adversarial attacks. It proves that AI relies on topological connectivity that humans don't register, creating a new class of 'invisible' security vulnerabilities.
From the abstract
Deep neural networks for 3D point cloud understanding have achieved remarkable success in object classification and recognition, yet recent work shows that these models remain highly vulnerable to adversarial perturbations. Existing 3D attacks predominantly manipulate geometric properties such as point locations, curvature, or surface structure, implicitly assuming that preserving global shape fidelity preserves semantic content. In this work, we challenge this assumption and introduce the first