[2110.11736] MANDERA: Malicious Node Detection in Federated Learning via Ranking

Computer Science > Machine Learning arXiv:2110.11736 (cs) [Submitted on 22 Oct 2021 (v1), last revised 26 Mar 2026 (this version, v3)] Title:MANDERA: Malicious Node Detection in Federated Learning via Ranking Authors:Wanchuang Zhu, Benjamin Zi Hao Zhao, Simon Luo, Tongliang Liu, Ke Deng View a PDF of the paper titled MANDERA: Malicious Node Detection in Federated Learning via Ranking, by Wanchuang Zhu and 4 other authors View PDF HTML (experimental) Abstract:Byzantine attacks hinder the deployment of federated learning algorithms. Although we know that the benign gradients and Byzantine attacked gradients are distributed differently, to detect the malicious gradients is challenging due to (1) the gradient is high-dimensional and each dimension has its unique distribution and (2) the benign gradients and the attacked gradients are always mixed (two-sample test methods cannot apply directly). To address the above, for the first time, we propose MANDERA which is theoretically guaranteed to efficiently detect all malicious gradients under Byzantine attacks with no prior knowledge or history about the number of attacked nodes. More specifically, we transfer the original updating gradient space into a ranking matrix. By such an operation, the scales of different dimensions of the gradients in the ranking space become identical. The high-dimensional benign gradients and the malicious gradients can be easily separated. The effectiveness of MANDERA is further confirmed by experimen...