Abstract: Since its inception, federated learning has experienced rapid development due to its distributed structure and advantages in privacy and security. However, the fairness issues arising from large-scale federated learning have affected the sustainability of federated learning systems. In response to the fairness issues in federated learning, recent research on fairness in federated learning has been systematically reviewed and deeply analyzed. Firstly, the workflow and definition of federated learning were explained, and biases and fairness concepts in federated learning were summarized. Secondly, commonly used datasets in fairness research in federated learning were detailed, and the challenges faced by fairness research were discussed. Finally, relevant research work was summarized from four aspects: data source selection, model optimization, contribution evaluation, and incentive mechanisms. The advantages, disadvantages, applicable scenarios, and experimental settings of these research methods were summarized, and the future research directions and trends in fairness in federated learning were anticipated.

Key words: Keywords: federated learning, fairness, data selection, model optimization, contribution evaluation, incentive mechanism

摘要： 联邦学习自诞生以来，凭借其分布式结构和隐私安全的优势，取得了快速发展。然而，大规模联邦学习所引发的公平性问题对联邦学习系统的可持续性造成了影响。针对联邦学习的公平性问题，对近年来联邦学习公平性的研究工作进行了系统梳理和深度分析。首先对联邦学习的工作流程和定义进行了解释，并总结了联邦学习中的偏见和公平性概念。其次详细归纳了联邦学习公平性研究中常用的数据集，并探讨了公平性研究所面临的挑战。最后从数据源选择、模型优化、贡献评估和激励机制四个方面归纳梳理了相关研究工作的优缺点、适用场景以及实验设置等，并展望了联邦学习公平性未来的研究方向和趋势。

关键词: 关键词: 联邦学习, 公平性, 数据选择, 模型优化, 贡献评估, 激励机制