体联网环境下的边缘智能协同隐私保护方案
Edge intelligent cooperative protection scheme under the Internet of Bodies environment
张海超 1李嘉坤 2刘东 3牛新征3
作者信息
- 1. 四川公安厅科技信息化总队,成都 610015
- 2. 电子科技大学信息与软件工程学院,成都 610054
- 3. 电子科技大学计算机科学与工程学院,成都 611731
- 折叠
摘要
针对现存智能模型,如医疗领域的智能诊断模型等存在准确率低、普适性较差的问题,同时用户在使用体联网设备提供的服务过程中存在隐私泄漏风险的问题,文中结合边缘计算和联邦学习,提出了一种体联网环境下的边缘智能协同隐私保护方案.设计了一种基于云-边-端分层的联邦学习训练模型,综合准确率较高,并提出了面向边云模型传输的差分隐私算法,能以较低的准确率损失换取强有力的隐私保护效果.该方案使用乳腺癌数据集在全连接网络上的仿真测试中效果良好,实验验证了该方法对隐私保护具有有效性.
Abstract
To solve the problems of low accuracy and poor universality of existing intelligent models,such as intelligent diagnosis models in the medical field,and the risk of privacy leakage when users use the serv-ices provided by the IoT devices,this paper proposes an edge intelligent collaborative privacy protection scheme under the IoT environment by combining edge computing and federated learning.A federated learn-ing training model based on cloud-edge-end layering is designed,and the comprehensive accuracy is high.A differential privacy algorithm for edge cloud model transmission is proposed,which can obtain strong pri-vacy protection effect with low accuracy loss.The scheme performs well in the simulation test using the breast cancer dataset on the fully connected network,and the experiment verifies the effectiveness of the method for privacy protection.
关键词
边缘计算/联邦学习/隐私保护/差分隐私/体联网Key words
edge computing/federal learning/privacy protection/differential privacy/Internet of Bodies引用本文复制引用
基金项目
四川省科技计划项目(2021YFS0391)
四川省重大科技专项(22DZX0046)
国家自然科学基金重点项目(61133016)
出版年
2024
NETL
NSTL
万方数据