TY - JOUR
T1 - TPC
T2 - 39th International Conference on Machine Learning, ICML 2022
AU - Chu, Wenda
AU - Li, Linyi
AU - Li, Bo
N1 - Funding Information:
The authors thank the anonymous reviewers for their valuable feedback. This work is partially supported by NSF grant No.1910100, NSF CNS No.2046726, C3 AI, and the Alfred P. Sloan Foundation. WC would like to thank the support from Institute for Interdisciplinary Information Sciences, Tsinghua University.
Publisher Copyright:
Copyright © 2022 by the author(s)
PY - 2022
Y1 - 2022
N2 - Point cloud models with neural network architectures have achieved great success and have been widely used in safety-critical applications, such as Lidar-based recognition systems in autonomous vehicles. However, such models are shown to be vulnerable to adversarial attacks that aim to apply stealthy semantic transformations such as rotation and tapering to mislead model predictions. In this paper, we propose a transformation-specific smoothing framework TPC, which provides tight and scalable robustness guarantees for point cloud models against semantic transformation attacks. We first categorize common 3D transformations into three categories: additive (e.g., shearing), composable (e.g., rotation), and indirectly composable (e.g., tapering), and we present generic robustness certification strategies for all categories respectively. We then specify unique certification protocols for a range of specific semantic transformations and their compositions. Extensive experiments on several common 3D transformations show that TPC significantly outperforms state of the art. For example, our framework boosts the certified accuracy against twisting transformation along the z-axis (within ±20◦) from 20.3% to 83.8%. Codes and models are available at https://github.com/Qianhewu/Point-Cloud-Smoothing.
AB - Point cloud models with neural network architectures have achieved great success and have been widely used in safety-critical applications, such as Lidar-based recognition systems in autonomous vehicles. However, such models are shown to be vulnerable to adversarial attacks that aim to apply stealthy semantic transformations such as rotation and tapering to mislead model predictions. In this paper, we propose a transformation-specific smoothing framework TPC, which provides tight and scalable robustness guarantees for point cloud models against semantic transformation attacks. We first categorize common 3D transformations into three categories: additive (e.g., shearing), composable (e.g., rotation), and indirectly composable (e.g., tapering), and we present generic robustness certification strategies for all categories respectively. We then specify unique certification protocols for a range of specific semantic transformations and their compositions. Extensive experiments on several common 3D transformations show that TPC significantly outperforms state of the art. For example, our framework boosts the certified accuracy against twisting transformation along the z-axis (within ±20◦) from 20.3% to 83.8%. Codes and models are available at https://github.com/Qianhewu/Point-Cloud-Smoothing.
UR - http://www.scopus.com/inward/record.url?scp=85163062956&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85163062956&partnerID=8YFLogxK
M3 - Conference article
AN - SCOPUS:85163062956
SN - 2640-3498
VL - 162
SP - 4035
EP - 4056
JO - Proceedings of Machine Learning Research
JF - Proceedings of Machine Learning Research
Y2 - 17 July 2022 through 23 July 2022
ER -