DOI：10.3969/j.issn.1672-3872.2023.23.019
作者：乔富强，周劲（武汉轻工大学电气与电子工程学院，湖北 武汉 430048）
摘 要：【目的】在碾米机生产线中，碾米机的数据隐私问题较为重要，利用联邦学习方法来保障碾米机生产线的数据安全。【方法】研究团队介绍了联邦学习系统的组成部分以及联邦学习分类，设计了一种基于隐私保护通信（PPTrans）的联邦学习技术，PPTrans通信包括碾米机生产线主体、智能网关和碾米机生产线之间的云服务器的处理流程。针对该通信技术，结合微众银行的FATE联邦学习框架和Docker容器技术来构建碾米机生产线的隐私保护系统。经过网关接收DHCP请求、转发设备信息、记录设备代理名称、处理FL训练计算请求等流程，来实现对碾米机生产线数据隐私的安全保护。【结果】该系统可以实现对碾米机内部工作数据的有效隔离，有效减少碾米机与对应云服务之间存在的用户原始数据隐私泄露的情况，具有一定的实用性。【结论】相较于传统的网络安全防御体系而言，使用联邦学习的数据隔离方法操作简单，工作量小，拥有较高的系统稳定性，适用于多种加工企业的网络安全设备。
关键词：联邦学习；网络安全；隐私保护；稻米加工
Defense Design of Federated Learning System Based on Rice Mill Control System
Qiao Fuqiang, Zhou Jin (School of Electrical and Electronic Engineering, Wuhan Polytechnic University, Hubei Wuhan 430048)
Abstract: [Objective] In the rice mill production line, the data privacy issue of the rice mill is more important, and the federated learning method is used to protect the data security of the rice mill production line. [Method] The research team introduced the components of the federated learning system and federated learning classification, and designed a federated learning technology based on privacy-preserving communication (PPTrans). PPTrans communication includes the processing process of the cloud server between the main body of the rice mill production line, the intelligent gateway and the rice mill production line. In response to this communication technology, a privacy protection system for the rice milling machine production line is constructed by combining WeBank’s FATE federated learning framework and Docker container technology. Through the process of receiving DHCP requests through the gateway, forwarding device information, recording device agent names, and processing FL training calculation requests, security protection of data privacy for the rice milling machine production line is achieved. [Result] This system can effectively isolate the internal working data of the rice milling machine, effectively reduce the leakage of user original data privacy between the rice milling machine and the corresponding cloud service, and has certain practicality. [Conclusion] Compared with the traditional network security defense system, the data isolation method using federated learning is simple to operate, has small workload, has high system stability, and is suitable for network security equipment of various processing enterprises. Keywords: federated learning; network security; privacy protection; rice processing
引文信息：[1]乔富强，周劲.基于碾米机控制系统的联邦学习系统防御设计[J].南方农机，2023，54(23)：78-81+120.
查看全文请下载PDF文件↓
