摘要
针对现有键值数据收集机制在数据精度估计方面的局限性,提出了一种基于本地差分隐私的迭代自适应划分键值数据收集方法。通过两阶段设计实现精准估计,第一阶段在本地对所有键值对值域进行初步划分并完成编码、扰动,由服务器聚合估计;第二阶段基于第一阶段估计结果自适应优化值域划分区间,迭代执行第一阶段步骤,利用误差阈值控制迭代次数,以实现准确的估计结果。理论分析了算法满足本地差分隐私和无偏估计。实验结果表明了算法的有效性和实用性。
To address the limitations of existing key-value data collection mechanisms in data accuracy estimation,a novel iterative adaptive partitioning key-value data collection method based on local differential privacy was proposed.Accurate estimation was achieved through a two-stage design.In Stage I,the value range of all key-value pairs was preliminarily divided,encoded,and perturbed locally,and then aggregated and estimated by the server.In Stage II,based on the estimation results of Stage I,the value range division intervals were adaptively optimized,and the steps of Stage I were iteratively executed,with the number of iterations controlled by an error threshold,to achieve accurate estimation results.Theoretical analysis demonstrates that the proposed algorithm satisfies local differential privacy and provides unbiased estimation.Experimental results verify the effectiveness and practicality of the proposed algorithm.
作者
孙庆毅
李晓会
兰洁
贾旭
李波
SUN Qing-yi;LI Xiao-hui;LAN Jie;JIA Xu;LI Bo(School of Electronics and Information Engineering,Liaoning University of Technology,Jinzhou 121001,China;College of Science,Liaoning University of Technology,Jinzhou 121001,China)
出处
《计算机工程与设计》
北大核心
2026年第1期154-164,共11页
Computer Engineering and Design
基金
国家自然科学基金项目(62203201、61802161)
辽宁省应用基础研究计划基金项目(2022JH2/101300278、2022JH2/101300279)
2024年辽宁省属本科高校基本科研业务费专项基金项目(LJZZ212410154025、LJZZ222410154004)
辽宁工业大学研究生教育改革创新基金项目(YJG2023013)。
关键词
本地差分隐私
键值数据
自适应划分
频率估计
均值估计
数据收集
隐私保护
local differential privacy
key-value data
adaptive partitioning
frequency estimation
mean value estimation
data collection
privacy protection
