SAFER-FL: A ROBUST CLOUD–EDGE FEDERATED LEARNING FRAMEWORK FOR SECURE COMMUNICATION AND COMPUTATION UNDER POISONING ATTACKS

The dataset proposes SAFER-FL, a strong and safe federated learning framework that solves the problems of secure communication and computation in cloud-edge settings that are under poisoning threats. The framework uses important tools, including accuracy-based client filtering, trimmed aggregation, and safe aggregation, to lessen the effects of bad client updates.

本数据集提出SAFER-FL——一款兼具高性能与安全性的联邦学习（Federated Learning）框架，可解决投毒攻击威胁下云边端环境中的安全通信与安全计算难题。该框架采用基于准确率的客户端筛选、裁剪聚合与安全聚合等关键技术，以削弱恶意客户端更新所带来的负面影响。