摘要
Federated learning (FL) allows multiple clients to train deep learning models collaboratively while protecting sensitive local datasets. However, FL has been highly susceptible to security for federated backdoor attacks (FBA) through injecting triggers and privacy for potential data leakage from uploaded models in practical application scenarios. FBA defense strategies consider specific and limited attacker models, and a sufficient amount of noise injected can only mitigate rather than eliminate the attack. To address these deficiencies, we introduce a Robust Federated Backdoor Defense Scheme (RFBDS) and Privacy preserving RFBDS (PrivRFBDS) to ensure the elimination of adversarial backdoors. Our RFBDS to overcome FBA consists of amplified magnitude sparsification, adaptive OPTICS clustering, and adaptive clipping. The experimental evaluation of RFBDS is conducted on three benchmark datasets and an extensive comparison is made with state-of-the-art studies. The results demonstrate the promising defense performance from RFBDS, moderately improved by 31.75% similar to 73.75% in clustering defense methods, and 0.03% similar to 56.90% for Non-IID to the utmost extent for the average FBA success rate over MNIST, FMNIST, and CIFAR10. Besides, our privacy-preserving shuffling in PrivRFBDS maintains is 7.83e-5 similar to 0.42x that of state-of-the-art works.
学科与关键词
- Breaking Distributed Backdoor Defenses for Federated Learning in Non-IID SettingsYang, Jijia*; Shu, Jiangang; Jia, Xiaohua18th IEEE International Conference on Mobility, Sensing and Networking (MSN), ELECTR NETWORK, 2022-12-14 To 2022-12-16.被引 1
- APFed: Anti-Poisoning Attacks in Privacy-Preserving Heterogeneous Federated LearningChen, Xiao; Yu, Haining*; Jia, Xiaohua; Yu, XiangzhanIEEE Transactions on Information Forensics and Security, 2023, 18: 5749-5761.被引 1
- Data and Model Poisoning Backdoor Attacks on Wireless Federated Learning, and the Defense Mechanisms: A Comprehensive SurveyWan, Yichen; Qu, Youyang; Ni, Wei; Xiang, Yong; Gao, Longxiang*; Hossain, EkramIEEE Communications Surveys & Tutorials, 2024, 26(3): 1861-1897.被引 6
