隐私保护频繁项集挖掘中的分组随机化模型  

Grouping Randomized Model in Privacy Preserving Frequent Item Set Mining

作　　者：郭宇红[1] 童云海 GUO Yuhong;TONG Yunhai(School of Information Science and Technology,University of International Relations,Beijing 100091,China;Department of Intelligence Science,Peking University,Beijing 100871,China)

机构地区：[1]国际关系学院信息科技学院,北京100091 [2]北京大学智能科学系,北京100871

出　　处：《华侨大学学报（自然科学版）》2020年第2期230-236,共7页Journal of Huaqiao University(Natural Science)

基　　金：国际关系学院中央高校基本科研业务费专项资金资助项目(3262017T48,3262018T02)。

摘　　要：通过对隐私保护频繁项集挖掘问题的研究,发现现有的单参数随机化回答模型调控的数据范围宽、粒度粗,导致无法实现精细化、差异化的隐私保护的问题.在沃纳模型、单参数等随机化模型的基础上,提出个体分组多参随机化PN/g模型,给出其在隐私保护频繁项集挖掘中的支持度重构方法.研究结果表明:该模型面向多样化、差异化的隐私保护需求,将N个不同个体分为若干组,每组设置不同的随机化参数,可实现差异化的隐私保护效果.实例分析表明:结合所提出的支持度重构方法,可实现基于分组随机化的隐私保护频繁项集挖掘,在保护不同群体隐私的同时,挖掘到有效的频繁项集与关联规则.Through the research of privacy preserving frequent item set mining,it is found that the existing single-parameter randomized response model regulates the data range wide and the granularity coarse,which leads to the problem that the privacy protection can not be refined and differentiated.Based on Warner model and single-parameter randomization model,an individual grouping multi-parameter randomized model of PN/g is proposed.The corresponding support degree reconstruction method in privacy preserving frequent item set mining is given.The research results show that the model is oriented to diversified and differentiated privacy protection needs,and N different individuals are divided into several groups,and each group is set with different randomization parameters,which can achieve differentiated privacy protection effects.Example analysis shows that combined with the proposed support reconstruction method,privacy preserving frequent item set mining based on grouping randomization can be realized,while protecting the privacy of different groups,effective frequent item sets and association rules can be mined.

关 键 词：随机化回答 隐私保护 频繁项集 支持度重构 数据挖掘 沃纳模型 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]