Robust Federated Learning against Backdoor Attackers
Robust Federated Learning against Backdoor Attackers
复制标题
DOI：
10.1109/infocomwkshps57453.2023.10225922
复制
发表时间：
2023-05
期刊：
影响因子：
--
通讯作者：
Priyesh Ranjan;Ashish Gupta;Federico Coró;Sajal Kumar Das
中科院分区：
文献类型：
--
作者：
Priyesh Ranjan;Ashish Gupta;Federico Coró;Sajal Kumar Das
联邦学习是一种保护隐私的分布式学习替代方案，不涉及数据传输。由于服务器对客户机的行为没有任何控制，一些对手可能会参与学习，将破坏引入底层模型。后门攻击者就是这样一种攻击者，他将触发模式注入数据以操纵特定子任务上的模型结果。这项工作旨在识别后门攻击者，并通过隔离其权重更新来减轻其影响。利用客户端梯度之间的相关性，我们提出了两种图论算法来区分攻击者和良性客户端。在分类任务下，实验结果表明，对于在目标图像的不同位置添加后门触发模式的攻击者，我们的算法是有效的和鲁棒的。结果还表明，当攻击者数量大于正常客户端时，我们的算法优于现有的方法。
Federated learning is a privacy-preserving alter-native for distributed learning with no involvement of data transfer. As the server does not have any control on clients' actions, some adversaries may participate in learning to introduce corruption into the underlying model. Backdoor attacker is one such adversary who injects a trigger pattern into the data to manipulate the model outcomes on a specific sub-task. This work aims to identify backdoor attackers and to mitigate their effects by isolating their weight updates. Leveraging the correlation between clients' gradients, we propose two graph theoretic algorithms to separate out attackers from the benign clients. Under a classification task, the experimental results show that our algorithms are effective and robust to the attackers who add backdoor trigger patterns at different location in targeted images. The results also evident that our algorithms are superior than existing methods especially when numbers of attackers are more than the normal clients.