Curvature-Invariant Adversarial Attacks for 3D Point Clouds
Authors: Jianping Zhang, Wenwei Gu, Yizhan Huang, Zhihan Jiang, Weibin Wu, Michael R. Lyu
AAAI 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Comprehensive experimental comparisons confirm the superiority of our approach. Notably, our strategy can achieve 7.2% and 14.5% improvements in Hausdorff distance and Gaussian curvature measurements of the imperceptibility. |
| Researcher Affiliation | Academia | 1Department of Computer Science and Engineering, The Chinese University of Hong Kong 2School of Software Engineering, Sun Yat-sen University |
| Pseudocode | Yes | Algorithm 1: Curvature-Invariant Method |
| Open Source Code | No | The paper does not provide a direct link to open-source code or explicitly state that the code for the described methodology is publicly available. |
| Open Datasets | Yes | Dataset. We follow the dataset selection of the baseline method (Huang et al. 2022) by utilizing the dataset Model Net40. Model Net40 consists of 12,311 CAD models from 40 object categories, in which 9,843 models are intended for training and the other 2,468 for testing. |
| Dataset Splits | No | The paper specifies the number of models for training and testing, but does not explicitly provide details for a separate validation split or the methodology for creating one. |
| Hardware Specification | Yes | All the experiments are conducted on a server equipped with one TITAN X GPU. |
| Software Dependencies | No | The paper mentions the use of deep learning models and common tasks but does not specify any software libraries or frameworks with version numbers (e.g., Python, PyTorch, TensorFlow versions). |
| Experiment Setup | Yes | Parameter. For a fair comparison, we set the maximum L budget of all the attacking methods to be ϵ = 0.16. In addition, the number of iterations is set to be T = 5, and the step length is 0.07. In the experiment, we adopt the untargeted attack under the same setting to evaluate the imperceptibility and attacking performance. For our approach, we set the hyper-parameter γ to regularize the gradient on the tangent plane to be 0.3. |