摘要
针对作战大数据联合共享建模和敏感数据保护的矛盾,对联邦学习框架展开研究,探讨了联邦学习模型产生的研究背景,模型原理和框架;分析了横向和纵向联邦学习模型的建模方法以及在联合作战数据共享领域的应用场景;并以装备保障作战数据服务领域为示范应用场景,介绍了基于联邦学习技术的数据共享交换典型应用;最后,从统计异构性、系统异构性、网络通信和安全问题等角度讨论联邦学习联合作战数据共享中面临的难题与挑战。
Aiming at the contradiction between joint sharing modeling of operation big data and sensitive data protection,the federated learning framework is studied,and the research background and model principle are discussed.Secondly,the modeling methods of horizontal and vertical federated learning models and the application scenarios in the field of combined operation data sharing are analyzed.And then the typical application of data sharing and exchange based on federated learning technology in the field of equipment support operation data service is introduced.Finally,the difficulties and challenges in data sharing of federated learning joint operations are discussed from the aspects of statistical heterogeneity,system heterogeneity,network communication and security.
作者
陈财森
纪伯公
黄辰
向阳霞
CHEN Cai-sen;JI Bo-gong;HUANG Chen;XIANG Yang-xia(Army Academy of Armored Forces,Beijing 100072,China)
出处
《装甲兵学报》
2022年第1期98-103,共6页
Journal of Armored Forces
基金
国家自然科学基金资助项目(U1836101)
军委科技委基础加强计划技术领域基金项目(2019-JCJQ-JJ-031)
军队科研计划项目
关键词
联邦学习
数据共享
隐私保护
共享建模
federated learning
data sharing
privacy protection
shared modeling
