Provable Defense Against Clustering Attacks on 3D Point Clouds
Lately, the literature on adversarial robustness spans from images to other domains such as point clouds. In this work, we consider clustering attacks on 3D point clouds and devise a provable defense mechanism to counter them. Specifically, we adopt a randomized smoothing strategy for 3D point clouds and derive a robustness certificate based on the cluster radius rather than the number of adversarial points. Our experiments on ModelNet40 and ScanObjectNN datasets using the PointNet classifier demonstrate the effectiveness of our defense mechanism against targeted and untargeted clustering attacks with a large number of adversarial points.
Code (0)
등록된 구현이 없습니다.
Tasks
Adversarial RobustnessClusteringMethods 이 논문이 사용한 방법론
Similar Papers 제목 키워드 기반
Deep Partition Aggregation: Provable Defense against General Poisoning Attacks
Adversarial poisoning attacks distort training data in order to corrupt the test-time behavior of a classifier. A provable defense provides a certificate for each test sample, which is a lower bound on the magnitude of a…
Deep Partition Aggregation: Provable Defenses against General Poisoning Attacks
Adversarial poisoning attacks distort training data in order to corrupt the test-time behavior of a classifier. A provable defense provides a certificate for each test sample, which is a lower bound on the magnitude of a…
Connecting Lyapunov Control Theory to Adversarial Attacks
Significant work is being done to develop the math and tools necessary to build provable defenses, or at least bounds, against adversarial attacks of neural networks. In this work, we argue that tools from control theory…
MathTextGuard: Provable Defense against Backdoor Attacks on Text Classification
Backdoor attacks have become a major security threat for deploying machine learning models in security-critical applications. Existing research endeavors have proposed many defenses against backdoor attacks. Despite demo…
ClassificationSentencetext-classificationText ClassificationProvable Defenses against Spatially Transformed Adversarial Inputs: Impossibility and Possibility Results
One intriguing property of neural networks is their inherent vulnerability to adversarial inputs, which are maliciously crafted samples to trigger target networks to misbehave. The state-of-the-art attacks generate adver…