The Technology Of Anonymity Research In Privacy Preserving Data Publishing

Human society has entered an information era today. All kinds of informationtechnology have made substantial progresses，producing large amounts of data. Thecollection and sharing of the data has brought great convenience to people. Datapublishing technology as an effective method for data sharing provides a strong supportfor the exchanging and sharing of data. However, with the development of datapublishing technology, there is a growing concern on the pirvacy information leakageproblem. How to effectively avoid the leakage of privacy information is becoming amajor challenges for data publishing. As a privacy protection technology in datapublishing，data anonymization has attracted much attention because of its simplicilty intheory and implementation. In this thesis, the main contents and contirbutions include:(1)Firstly, in correspondence to several serious pirvacy information leakageincidents in recent years, we introduce the research status of the anonymizationtechnique from the three aspects of micro data，social networks and hypergraphsrespectively. And then we descirbe a variety of attack methods contrapose pirvacyinformation and different anonymization techniques for pirvacy preserving datapublishing. Moreover, we introduce the operation mechanism of k-anonymity model indetail, discuss the advantages and disadvantages of the model, and describe theinformation loss metircs. In addition，we introduce several improvements on typicalanonymity models and summarize the attacks can be dealt with by them.(2)For the micro data with multi-dimensional numerical sensitive attirbutesreleased, we present MNSAGM anonymity model and algorithm, and it can againstproximity breach which is a characteirstic of numerical sensitive data. The model isbased on k-anonymity model. First，it divides the values of numeircal sensitiveattributes of each dimension into approximate groups, that can be adjusted by settingdifferent threshold values8. Then，it builds a multi-dimensional barrel and selectsappropirate records to form record groups. Finally，it generalizes the quasi-identifiers ineach record groups and get the anonymity table. The expeirmental results show that thealgorithm can effectively resist approximate attack to micro data with multi?dimensional numeircal sensiitve attirbutes.(3)For pirvacy preserving data publishing of hypergraphs. Firstly，we introduce therelated knowledge of hypergraphs, and put forward the concept of hypergraph Laplacesequence sets based on hypergraph signless Laplace matrix. Then we point out that the Laplace sequence attack of a hypergraph may lead to identity disclosure. Finally, inorder to cope with the hypergraph Laplace sequence attack, the paper puts forward theconcept of hypergraph anonymity Laplace sequence and the anonymity model whichcan resist this kind of attacks. In order to realize the model, we propose a two-stepapproximate algorithm: the first step makes the hypergraph Laplace sequence satisfiesanonymity, and the second step constructs a new hypergraph based on the Laplacesequence sets. The expeirmental results show that the algorithm can effectively resistthe Laplace sequence attack in hypergraphs.