A Privacy-Preserving Method Of Transactional Data Based On Slicing Technology

With ever expanding Internet, all daily activities of human beings, such as online/offline shopping, medical treatment, web search queries and even daily communication (to name a few) can be traced, recorded and stored in the transactional dataset. Although publishing and sharing these data implying massive information can provide powerful data support for various research and business decisions, it simultaneously poses severe threats to individual privacy. Therefore, no matter in theory or in reality, it is quite important to put forward an effective method to preserve users’ privacy and reduce information loss as far as possible while publishing and sharing the transactional dataset.From the perspective of the owner of a transactional dataset, in this paper we comprehensively consider membership privacy protection, identity privacy protection, attribute privacy protection and association rule protection, and propose a novel utility metric named as utility gain. To make a better tradeoff between privacy protection and data utility, we propose two different privacy-preserving methods for transactional data based on slicing technology which are entropy L-diversity slicing method and t-closeness slicing method.Firstly, in this paper we put forward the entropy L-diversity slicing method for transactional data based on the entropy L-diversity model and slicing technology to overcome the existing shortcomings of the privacy-preserving research of transactional data. Based on fundamental ideas of the method, we give a specific algorithm design and prove high efficiency of the method. We conduct numeric experiments on three kinds of transactional datasets. And the experimental results demonstrate that the entropy L-diversity slicing method can not only effectively protect membership privacy, identity privacy, attribute privacy and association rules, but also outperform the entropy L-diversity anatomy method in reducing utility loss.Secondly, considering the problem of skewness attack that the entropy L-diversity slicing method is insufficient for, in this paper we introduce the t-closeness model into the privacy-preserving research of transactional data, and propose the t-closeness slicing method for transactional data based on slicing technology. According to the basic idea of the method, we design a specific algorithm and analyze its time complexity. The experimental results show that the t-closeness slicing method can not only effectively protect membership privacy, identity privacy, attribute privacy and association rules, but also outperform the t-closeness anatomy method in preserving data utility by numerical experiments.In this paper we propose privacy-preserving methods for transactional data to make a more effective tradeoff between privacy protection and data utility. That is, on the premise of effectively protecting personal privacy, we improve the usability of the released transactional dataset as far as possible.