An efficient online histogram publication method for data streams with local differential privacy  

作　　者：Tao TAO Funan ZHANG Xiujun WANG Xiao ZHENG Xin ZHAO 

机构地区：[1]School of Computer Science and Technology,Anhui University of Technology,Maanshan 243032,China [2]Anhui Engineering Research Center for Intelligent Applications and Security of Industrial Internet,Maanshan 243032,China [3]Engineering Research Institute,Anhui University of Technology,Maanshan 243032,China [4]Shengli No.1 Middle School of Dongying City,Dongying 257000,China

出　　处：《Frontiers of Information Technology & Electronic Engineering》2024年第8期1096-1109,共14页信息与电子工程前沿（英文版）

基　　金：supported by the Anhui Provincial Natural Science Foundation,China(Nos.2108085MF218 and 2022AH040052);the University Synergy Innovation Program of Anhui Province,China(No.GXXT-2023-021);the Key Program of the Natural Science Foundation of the Educational Commission of Anhui Province of China(No.2022AH050319);the National Natural Science Foundation of China(Nos.62172003 and 61402008)。

摘　　要：Many areas are now experiencing data streams that contain privacy-sensitive information.Although the sharing and release of these data are of great commercial value,if these data are released directly,the private user information in the data will be disclosed.Therefore,how to continuously generate publishable histograms(meeting privacy protection requirements)based on sliding data stream windows has become a critical issue,especially when sending data to an untrusted third party.Existing histogram publication methods are unsatisfactory in terms of time and storage costs,because they must cache all elements in the current sliding window(sW).Our work addresses this drawback by designing an efficient online histogram publication(EOHP)method for local differential privacy data streams.Specifically,in the EOHP method,the data collector first crafts a histogram of the current SW using an approximate counting method.Second,the data collector reduces the privacy budget by using the optimized budget absorption mechanism and adds appropriate noise to the approximate histogram,making it possible to publish the histogram while retaining satisfactory data utility.Extensive experimental results on two different real datasets show that the EOHP algorithm significantly reduces the time and storage costs and improves data utility compared to other existing algorithms.

关 键 词：Data stream Differential privacy Sliding windows Approximate counting 

分 类 号：TP309[自动化与计算机技术—计算机系统结构]