基于差分隐私的数据流频繁项集发布  

Dataflow frequent item set publishing based on differential privacy

作　　者：张鸿鸣[1] 鲍晓涵 倪巍伟[2] ZHANG Hong-ming;BAO Xiao-han;NI Wei-wei(Smart Grid Service Center,Jiangsu Frontier Electric Technology Co.,LTD,Nanjing 210000,China;School of Computer Science and Engineering,Southeast University,Nanjing 211189,China)

机构地区：[1]江苏方天电力技术有限公司智能电网服务中心,江苏南京210000 [2]东南大学计算机科学与工程学院,江苏南京211189

出　　处：《计算机工程与设计》2022年第11期3051-3056,共6页Computer Engineering and Design

基　　金：国家自然科学基金项目(61772131)。

摘　　要：针对数据流频繁项集发布中的隐私保护问题,提出隐私保护数据流频繁项集发布方法DP_DFIM。设计自适应的w-动态滑动窗口协议,实现发布精度独立于w值;利用频繁项信息进行事务截断,引入负项概念,最大化保留事务中的频繁项集,降低截断误差;在此基础上,提出隐私保护数据流频繁项集发布方法DP_DFIM,兼顾数据流隐私和数据流频繁项集可用性。设计实验验证DP_DFIM算法发布的项集能在满足差分隐私的前提下保证项集及其计数的准确性。To solve the problem of privacy leakage in frequent item set publishing of data stream,a method of privacy protection for frequent item set publishing of data stream DP_DFIM was proposed.An adaptive w-dynamic sliding window protocol was designed to achieve publishing accuracy independent of w value.Frequent item information was used to truncate transactions,and the concept of negative item was introduced to maximize the frequent item set in reserved transactions and reduce truncation error.On this basis,a privacy-protected frequent item set publishing method,DP_DFIM was proposed,to realize the secure release of frequent item set of data stream.Theoretical analysis and experimental results show that the itemset published by the proposed algorithm can guarantee the accuracy of itemset and its counting on the premise of satisfying differential privacy.

关 键 词：数据流发布 频繁项集发布 差分隐私 w-滑动窗口协议 事务截断 

分 类 号：TP309.2[自动化与计算机技术—计算机系统结构]