摘要
目前关于隐私保护数据发布的研究大多是面向低维的关系型数据,其相关模型及算法无法直接用于解决稀疏的高维事务型数据发布中可能存在的隐私泄露问题.本文以剖分技术为基础,设计出一个面向隐私保护事务型数据发布的p-剖分l-多样化匿名算法.算法通过计算事务型数据中属性间的均方列联系数将高维属性集剖分成互不相交的p个属性子集,而后对事务型数据进行记录划分,使记录划分后的事务型数据关于p个属性子集满足l-多样化的要求.实验对匿名前后事务型数据的关联规则挖掘结果进行比较分析.理论分析和实验结果表明,本文的算法可安全地实现事务型数据发布的隐私保护,同时保证发布数据的可用性较高.
Recently,privacy preserving data publishing has been a hot topic in data privacy preserving research fields.Exist research on privacy preserving data publishing mainly focuses on relational data with low dimensionality.However,many applications require privacy preserving publishing of transactional data,which has no structure and can be extremely high dimensional.Furthermore,unlike most previous works on relational data publishing,it is hard to distinguish transactional data as sensitive and non-sensitive,which makes traditional models and methods unusable.In this paper,we consider all the combination of itemsets in transactional data as potential quasi-identifiers and potential sensitive data,depending on the point of view of the adversary.Inspired by the anatomy technique,we propose a p-anatomy l-diversity algorithm for privacy preserving transactional data publishing.The algorithm firstly anatomizes the attribute set of the transactional data into p disjoint subsets by calculating the mean-square contingency coefficient between attributes,and then partitions the tuples of the transactional data into some equivalence classes,each of which satisfies the l-diversity requirement with respect to the above p attribute subsets.Experimental analysis is designed by comparing the rule number and the accuracy of association rules mining on the transactional data before and after publishing.The theoretical analysis and experimental results show that our algorithm can safely preserve the privacy in transaction data publication,while ensuring high utility of the released data.
出处
《南京大学学报（自然科学版）》
CAS
CSCD
北大核心
2011年第5期551-558,共8页
Journal of Nanjing University（Natural Science）
基金
国家自然科学基金(61003057)
福建省自然科学基金(2010J01330)
福州大学科技发展基金(2010-XY-20)
