Robustness Certification for Point Cloud Models

TL;DR

This paper introduces 3DCertify, a novel verifier that certifies the robustness of 3D point cloud models against various transformations, enhancing safety in critical applications like autonomous driving.

Contribution

We present 3DCertify, the first scalable verifier for point cloud models that handles diverse transformations using novel relaxation techniques.

Findings

Certifies robustness against rotations up to ±60° for 95.7% of point clouds.

Max pool relaxation improves certification coverage by up to 15.6%.

Effective for both classification and segmentation tasks.

Abstract

The use of deep 3D point cloud models in safety-critical applications, such as autonomous driving, dictates the need to certify the robustness of these models to real-world transformations. This is technically challenging, as it requires a scalable verifier tailored to point cloud models that handles a wide range of semantic 3D transformations. In this work, we address this challenge and introduce 3DCertify, the first verifier able to certify the robustness of point cloud models. 3DCertify is based on two key insights: (i) a generic relaxation based on first-order Taylor approximations, applicable to any differentiable transformation, and (ii) a precise relaxation for global feature pooling, which is more complex than pointwise activations (e.g., ReLU or sigmoid) but commonly employed in point cloud models. We demonstrate the effectiveness of 3DCertify by performing an extensive evaluation on a wide range of 3D transformations (e.g., rotation, twisting) for both classification and part segmentation tasks. For example, we can certify robustness against rotations by $\pm$60{\deg} for 95.7% of point clouds, and our max pool relaxation increases certification by up to 15.6%.