基于本地差分隐私的医疗数据收集方法  

Medical data collection algorithm based on local differential privacy

作　　者：王金鹏 李晓会[1] 贾旭[1] WANG Jin-peng;LI Xiao-hui;JIA Xu(School of Electronics and Information Engineering,Liaoning University of Technology,Jinzhou 121001,China)

机构地区：[1]辽宁工业大学电子与信息工程学院,辽宁锦州121001

出　　处：《计算机工程与设计》2024年第10期2929-2935,共7页Computer Engineering and Design

基　　金：国家自然科学基金青年基金项目(61802161);辽宁省应用基础研究计划基金项目(2022JH2/101300278、2022JH2/101300279);辽宁工业大学研究生教育改革创新基金项目(YJG2023013)。

摘　　要：针对现有医疗数据收集算法无法有效抵抗背景知识攻击和不可信第三方的隐私泄露问题,提出一种基于本地差分隐私的医疗数据收集方法。设计基于Count-Min Sketch和GRR算法的两阶段数据收集框架,利用随机采样技术避免隐私预算分割,降低数据收集的通信代价和噪声误差,通过对高低频症状分别抽样扰动收集统计,降低数据哈希冲突导致的误差问题。理论分析算法满足本地差分隐私。实验结果表明,该方法频率估计的精确度、运行时间和通信开销优于对比方法。To address the issues of existing medical data collection algorithms being unable to effectively resist background know-ledge attacks and the lack of a trusted third party leading to privacy breaches,a medical data collection method was proposed.A two-stage data collection framework based on the Count-Min Sketch technique and GRR algorithm was utilized.Random sampling techniques were employed to avoid privacy budget fragmentation,communication costs and noise errors were reduced.High and low frequency symptoms were separately sampled and perturbed to mitigate errors caused by hash collision.Theoretical analysis indicates that the proposed method meets the local differential privacy requirement.Experimental results indicate that the algorithm outperforms the comparative methods in both frequency estimation accuracy,runtime consumption and communication expense.

关 键 词：医疗数据收集 本地差分隐私 草图结构 分层收集 不可信第三方 隐私保护 数据可用性 

分 类 号：TP309.2[自动化与计算机技术—计算机系统结构]