README.md

awesome-3D-point-cloud-attacks

List of state of the art papers, code, and other resources focus on 3D point cloud attacks and defense

Attacks

• Extending Adversarial Attacks and Defenses to Deep 3D Point Cloud Classifiers.

  • [Code]

• Adversarial point perturbations on 3D objects. ECCV 2020

  • [Code]

• PointCloud Saliency Maps. ICCV 2019 drop points attack

  • [Code]

• LG-GAN: Label Guided Adversarial Network for Flexible Targeted Attack of Point Cloud Based Deep Networks. CVPR 2020

  • [Code]

• Robust Adversarial Objects against Deep Learning Models. AAAI 2020 knn attack

  • [Code]

• Self-Robust 3D Point Recognition via Gather-Vector Guidance. CVPR 2020 FGM IFGM MIFGM PGD

• On Isometry Robustness of Deep 3D Point Cloud Models Under Adversarial Attacks. CVPR 2020

  • [Code]

• Adversarial Autoencoders for Generating 3D Point Clouds.

• Generating 3D Adversarial Point Clouds. CVPR 2019 CW attack

  • [Code]

• PU-GAN: a Point Cloud Upsampling Adversarial Network.

  • [Code]

• 3D Point Cloud Generative Adversarial Network Based on Tree Structured Graph Convolutions.

  • [Code]

• Adversarial Autoencoders for Compact Representations of 3D Point Clouds.

  • [Code]

• Physically Realizable Adversarial Examples for LiDAR Object Detection. CVPR 2020

• AdvPC: Transferable Adversarial Perturbations on 3D Point Clouds. ECCV 2020

  • [Code]

• ShapeAdv: Generating Shape-Aware Adversarial 3D Point Clouds.

  • [Code]

• Robustness of 3D Deep Learning in an Adversarial Setting. CVPR 2019

  • [Code]

• Nudge Attacks on Point-Cloud DNNs.

• Generating Adversarial Surfaces via Band‐Limited Perturbations

• Geometry-aware generation of adversarial point clouds. TRAMI

  • [Code]

• Efficient Joint Gradient Based Attack Against SOR Defense for 3D Point Cloud Classification. ACM MM

• Adversarial Objects Against LiDAR Based Autonomous Driving Systems.

• Minimal Adversarial Examples for Deep Learning on 3D Point Clouds.

• Towards Robust LiDAR-based Perception in Autonomous Driving: General Black-box Adversarial Sensor Attack and Countermeasures.

• Geometric Adversarial Attacks and Defenses on 3D Point Clouds. autoencoder attack

  • [Code]

• Towards Universal Physical Attacks On Cascaded Camera-Lidar 3D Object Detection Models.

• Object Removal Attacks on LiDAR-based 3D Object Detectors.

• On the Adversarial Robustness of 3D Point Cloud Classification.

• Exploring Adversarial Robustness of Multi-Sensor Perception Systems in Self Driving.

• Fooling LiDAR Perception via Adversarial Trajectory Perturbation.

  • [Code]

• PointBA: Towards Backdoor Attacks in 3D Point Cloud.

• 3D Adversarial Attacks Beyond Point Cloud. mesh attack

  • [Code]

• Explainability-Aware One Point Attack for Point Cloud Neural Networks. one point attack

  • [Code]

• Adversarial Attack by Limited Point Cloud Surface Modifications.

• PointBA: Towards Backdoor Attacks in 3D Point Cloud. ICCV 2021

• A Backdoor Attack Against 3D Point Cloud Classifiers. ICCV 2021

• Generating Unrestricted 3D Adversarial Point Clouds.

• Local Aggressive Adversarial Attack of 3D point Cloud. ACML oral

  • [Code]

• Imperceptible Transfer Attack and Defense on 3D Point Cloud Classification.

• Attacking Point Cloud Segmentation with Color-only Perturbation segmentation

• Dynamics-aware Adversarial Attack of 3D Sparse Convolution Network segmentation

• Invisible for both Camera and LiDAR: Security of Multi-Sensor Fusion based Perception in Autonomous Driving Under Physical-World Attacks. IEEE S&P

  • [Code]

• Boosting 3D Adversarial Attacks with Attacking On Frequency AOF attack

  • [Code]

• Shape-invariant 3D Adversarial Point Clouds. CVPR 2022

  • [Code]

• Shape Prior Guided Attack: Sparser Perturbations on 3D Point Clouds AAAI 2022

• Improving transferability of 3D adversarial attacks with scale and shear transformations SS attack

• Isometric 3D Adversarial Examples in the Physical World NeurIPS 2022

• Rethinking Perturbation Directions for Imperceptible Adversarial Attacks on Point Clouds IEEE Internet of Things Journal

Defenses

• The art of defense: letting networks fool the attacker.

  • [Code]

• DUP-Net: Denoiser and Upsampler Network for 3D Adversarial Point Clouds Defense.

• IF-Defense: 3D Adversarial Point Cloud Defense via Implicit Function based Restoration.

  • [Code]

• PointCutMix: Regularization Strategy for Point Cloud Classification.

  • [Code]

• PointGuard: Provably Robust 3D Point Cloud Classification. CVPR 2021

• Defense-pointnet: Protecting pointnet against adversarial attacks.

• LPF-Defense: 3D Adversarial Defense based on Frequency Analysis.

  • [Code]

• Robust Structured Declarative Classifiers for 3D Point Clouds: Defending Adversarial Attacks with Implicit Gradients. CVPR 2022

• Improving Adversarial Robustness of 3D Point Cloud Classification Models ECCV 2022

  • [Code]

• Ada3Diff: Defending against 3D Adversarial Point Clouds via Adaptive Diffusion