基于联邦多任务学习的智慧教室隐私保护研究  

Research on Privacy Protection in Smart Classrooms Based on Federated Multi-task Learning

作　　者：杨雨浓 徐澳 张春炯 谢涛 YANG Yu-Nong;XU Ao;ZHANG Chun-Jiong;XIE Tao(College of Computer and Information Science,Chongqing Normal University,Chongqing,China 401331;AI Convergence Network Department,Asia University,Suwon,South Korea 16499;Faculty of Education,Southwest University,Chongqing,China 400715)

机构地区：[1]重庆师范大学计算机信息科学学院,重庆401331 [2]亚洲大学AI收敛网络系,韩国水原16499 [3]西南大学教育学部,重庆400715

出　　处：《现代教育技术》2024年第9期123-132,共10页Modern Educational Technology

基　　金：2022年国家自然科学基金项目“面向多视图教育数据的共享簇结构挖掘”(项目编号:62277043);2024年重庆市教委科学技术研究项目“数字教育场景下隐私应用与可信学习研究”(项目编号:KJZD-K202300515)的阶段性研究成果。

摘　　要：在5G时代,随着联网设备的增多,智慧教室中的数据传输变得异常复杂,将联邦学习应用于智慧教室的隐私保护具有重要意义。尽管已有的联邦学习技术在平衡公平性和准确性方面取得了积极进展,但在解决智慧教室场景中的隐私保护问题方面还不够彻底。基于此,文章首先剖析了智慧教室中的多任务学习场景,建立了适用于智慧教室的多任务机器学习公平性模型。随后,文章提出了基于联邦学习框架的智慧教室隐私保护策略——联邦多任务机器学习框架。为了验证此框架的有效性,文章开展了计算机仿真实验,实验结果显示:联邦多任务机器学习框架不仅实现了多任务机器学习公平性和准确性的平衡,而且具有较强的隐私保护能力。文章的研究可为教育数据隐私保护提供重要参考,并有力推动教育数字化转型。In the 5G era,with the increase of networked devices,data transmission in smart classrooms has become extremely complex,and it is of significant importance to apply federated learning to the privacy protection of smart classrooms.Although existing federated learning techniques have made positive progress in balancing fairness and accuracy,they have not gone far enough in the aspect of addressing privacy protection questions in smart classroom scenarios.Based on this,the paper firstly analyzed the multi-task learning scenarios in smart classrooms and established a multi-task machine learning fairness model suitable for the smart classrooms.Subsequently,the privacy protection strategy for smart classrooms under the federated learning framework was proposed,namely the federated multi-task machine learning framework.In order to verify the effectiveness of this framework,the computer simulation experiments were carried out in this paper.The experimental results showed that the federal multi-task machine learning framework not only achieved a balance between fairness and accuracy in multi-task machine learning,but also had strong privacy protection capabilities.The research of this paper could provide important reference for the privacy protection of educational data and strongly propel the digital transformation in education.

关 键 词：智慧教室 联邦学习 隐私保护 多任务机器学习 

分 类 号：G40-057[文化科学—教育学原理]