密码学报（中英文）2025，Vol.12Issue(3)：499-512,14.DOI:10.13868/j.cnki.jcr.000777
基于混洗差分隐私的键值对数据收集机制
Key-Value Pair Data Collection Mechanism Based on Shuffling Differential Privacy
摘要
Abstract
Key-value pair is a popular NoSQL data type where the key is an index of the value and the value is the specific value corresponding to the key.In practical applications,databases need to collect key-values to estimate frequencies,averages,etc.of data,but this process may compromise user privacy,and differential privacy mechanisms can help for this purpose,but may affect the accuracy of statistical results.Differential privacy mechanisms for key-value pairs,in contrast to ones for single data,need to maintain the correlation between keys and values when perturbing the data,which poses a great challenge for the design of related schemes.The existing schemes are based on localized differential privacy models,which are less accurate and have problems such as more iterations and difficult parameter selection.The shuffle model adds the shuffle operation between the user side and the server side of the localization model to separate the user's personal information from the data,which can achieve higher accuracy with the same degree of privacy protection.This study designs a key-value pair data collection mechanism by introducing the shuffle model,which enables the simultaneous estimation of the frequency of keys and the average value of each key.Through careful algorithm design,this mechanism ensures the accuracy of the final estimation results with the premise of differential privacy.关键词
差分隐私/键值对/混洗模型Key words
differential privacy/key-value/shuffle model分类
信息技术与安全科学引用本文复制引用
唐鸿宇,陈原,郭怡阳..基于混洗差分隐私的键值对数据收集机制[J].密码学报（中英文）,2025,12(3):499-512,14.基金项目
智能汽车安全技术全国重点实验室开放基金(IVSTSKL-202443)Open Fund of State Key Laboratory of Intelligent Vehicle Safety Technology(IVSTSKL-202443) （IVSTSKL-202443）
