Yu Feng, Dingxin Zhang, Dongnan Liu, Runkai Zhao, Yong Xia, Heng Huang, Weidong Cai
Arxiv 2025
We propose SPBA, a stealthy patch-wise backdoor attack for 3D point clouds that leverages local spectral perturbations guided by geometric complexity to achieve high attack success rates while maintaining state-of-the-art imperceptibility.
Yu Feng, Dingxin Zhang, Dongnan Liu, Runkai Zhao, Yong Xia, Heng Huang, Weidong Cai
Arxiv 2025
We propose SPBA, a stealthy patch-wise backdoor attack for 3D point clouds that leverages local spectral perturbations guided by geometric complexity to achieve high attack success rates while maintaining state-of-the-art imperceptibility.
Yu Feng*, Benteng Ma*, Dongnan Liu, Yanning Zhang, Weidong Cai, Yong Xia (* equal contribution)
IEEE Transactions on Image Processing (TIP) 2025
We propose Contrastive Neuron Pruning (CNP), a label-free defense that leverages contrastive learning to identify and prune backdoor-associated neurons, effectively mitigating backdoor attacks with minimal impact on model structure.
Yu Feng*, Benteng Ma*, Dongnan Liu, Yanning Zhang, Weidong Cai, Yong Xia (* equal contribution)
IEEE Transactions on Image Processing (TIP) 2025
We propose Contrastive Neuron Pruning (CNP), a label-free defense that leverages contrastive learning to identify and prune backdoor-associated neurons, effectively mitigating backdoor attacks with minimal impact on model structure.
Benteng Ma*, Yu Feng*, Geng Chen, Changyang Li, Yong Xia (* equal contribution)
Pattern Recognition (PR) 2023
We propose a novel Federated Adaptive Reweighting (FedAR) algorithm for medical image classification. FedAR employs a flexible re-weighting scheme that can balance adaptively the contributions of the amount of data and the performance of the local model on each client to the weight of that client.
Benteng Ma*, Yu Feng*, Geng Chen, Changyang Li, Yong Xia (* equal contribution)
Pattern Recognition (PR) 2023
We propose a novel Federated Adaptive Reweighting (FedAR) algorithm for medical image classification. FedAR employs a flexible re-weighting scheme that can balance adaptively the contributions of the amount of data and the performance of the local model on each client to the weight of that client.
Yu Feng*, Benteng Ma*, Jing Zhang, Shanshan Zhao, Yong Xia, Dacheng Tao (* equal contribution)
Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) 2022
We propose a novel Frequency-Injection based Backdoor Attack method (FIBA) that is capable of delivering attacks in various medical image analysis tasks.
Yu Feng*, Benteng Ma*, Jing Zhang, Shanshan Zhao, Yong Xia, Dacheng Tao (* equal contribution)
Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) 2022
We propose a novel Frequency-Injection based Backdoor Attack method (FIBA) that is capable of delivering attacks in various medical image analysis tasks.