摘要
本地差分隐私(local differential privacy, LDP)可以对可穿戴装置(wearable devices)采集到的数据进行隐私保护,每个用户都会在本地扰乱自己的数据,并且将扰动后的数据发送给数据汇聚服务器,以保护用户免受私人信息泄漏的影响。可穿戴装置采集到的数据是多维的,但是现有的针对可穿戴装置多维数据的个性化本地差分隐私保护研究比较少而且不完善。针对现有个性化本地隐私方案存在的最坏情况下噪声方差大的问题,采用结合机制,结合随机响应机制和分段机制,对数值型数据进行扰动,提出了一种处理数值型数据的个性化本地差分隐私保护方案,并将该方案应用到多维数值型数据,通过随机采样提高数据可用性。此外,分别从理论分析和仿真验证的角度对提出的本地差分隐私方案与现有解决方案进行了对比分析和实验。实验结果表明,提出的方案在最坏情况下的噪声方差方面优于现有解决方案,并且具有更好的数据可用性。
Local differential privacy(LDP) can protect the privacy of data collected by wearable devices. Each user will disturb his own data locally and send the disturbed data to the data aggregation server, to protect users from the impact of private information leakage. The data collected by the wearable device is multi-dimensional, but the existing research on personalized local differential privacy protection for the multi-dimensional data of the wearable device is relatively few and incomplete. Aiming at the problem of large noise variance in the worst case of the existing personalized local privacy schemes, the combination mechanism, combined with the random response mechanism and the piecewise mechanism, is used to disturb the numerical data, and a personalized local differential privacy protection scheme for processing numerical data is proposed. The scheme is applied to multi-dimensional numerical data, and data availability is improved through random sampling. In addition, comparative analysis and experiments are carried out on the proposed local differential privacy scheme and the existing solutions from the perspectives of theoretical analysis and simulation verification. Experimental results show that the proposed scheme is better than existing solutions in terms of noise variance in the worst case, and has better data availability.
作者
卢岑
沈苏彬
LU Cen;SHEN Su-bin(School of Internet of Things,Nanjing University of Posts and Telecommunications,Nanjing 210003,China;School of Computer,Nanjing University of Posts and Telecommunications,Nanjing 210023,China)
出处
《计算机技术与发展》
2022年第2期107-113,共7页
Computer Technology and Development
基金
国家自然科学基金(61502246)
南京邮电大学科研项目(NY220202)。
关键词
本地差分隐私
个性化
多维数据
可穿戴装置
随机响应
local differential privacy
personalization
multidimensional data
wearable devices
random response
