The Research On Privacy Preserving In Re-publication Of Dynamic Set-valued Data

With the rapid development and wide application of computer, database andnetwork information technology, more and more data are publicly available in thenetwork. Advances of data mining and many other technologies help people make useof the published data effectively, from which extract potential and valuable knowledge,and also promoted the development of data publication. However, published data maycontain a large number of personal privacy and sensitive information, which also facethe threat of privacy disclosure. Therefore, Research on privacy preserving has animportant role and significance in data publication. Set-valued data is one ofimportant types in data publication; there are also continuous insertion, deletion andmodification on the dataset in real applications, thus dataset need to be updated andre-published. In this dissertation, re-publication of dynamic set-valued data isconsidered as the research object, how to implement privacy protection, completeanonymous publication effectively, and better retain integrity and availability of thepublished result was studied on demand for re-publishing general formed dynamicset-valued data in general form.For privacy preserving in dynamic set-valued data re-publication, a privacypreserving model is presented and designed in this dissertation. The model is anorganic whole composed of data collection and updating, solution of privacypreserving, anonymization and data publication. It can achieve the goal ofimplementing privacy protection in re-publication of dynamically changing set-valueddataset.For problems in the existing set-valued data anonymous privacy preservingmethods, an improved k-anonymous algorithm is proposed and designed in thisdissertation. By integrating local recoding generalization with suppression techniques,the top-down partition based generalization algorithm is extended, employinggeneralization and suppression process on set-valued data in multiple rounds andtwo-phase, which result in effectively reducing information loss of the anonymousresult. Complete quality metrics are presented to verify promotion of the anonymousresult’s quality, which also provide a basis for the study of re-publication method.For re-publication of dynamic set-valued data, k-preserving principle is proposedbased on extended Sensitive attribute Update Graph theory for set-valued data.Privacy can be protected from being exposed by continuous employing transactionalk-anonymity on every single anonymization and maintaining diversity and continuityof the sensitive elements during re-publishing. Combined with the improvedk-anonymous algorithm, complete re-publication algorithm is also proposed, which isthe core content of the model proposed in this dissertation. On the basis of the research above, the privacy preserving model for dynamicset-valued data re-publication is implemented, on which further experiments areconducted on the core algorithms and re-publication methods by using real-worlddatasets. The results show that the model is feasible and effective; it achieved the goalof privacy preserving and improvement of the published data’s quality.