Federated learning framework based on trimmed mean aggregation rules

作者：Tianxiang Wang, Zhonglong Zheng, Feilong Lin · 发表于：Expert systems with applications · 年份：2025 · DOI：10.2139/ssrn.4181353 · 被引用次数：111 · 研究领域：Computer Science

This paper studies the problem of information security in the distributed learning framework. In particular, we consider the clients will always be attacked by Byzantine nodes and poisoning in the federated learning. Typically, aggregation rules are utilized to protect the model from the attacks in federated learning. However, classical aggregation methods such as Krum(·) and Mean(·) are not capable enough to deal with Byzantine attacks in which general deviations and multiple clients are attacked at the same time. We propose new aggregation rules, Tmean(·), to the federated learning algorithm, and propose a federated learning framework based on Byzantine resilient aggregation algorithm. Our Tmean(·) rules are derived from Mean(·) by appropriately trimming some of the values before averaging them. Theoretically, we provide theoretical analysis and understanding of Tmean(·). Extensive experiments validate the effectiveness of our approaches.