Research On T-closeness Privacy Protection Model With The Support Of Rough Set And Cluster

With the development of information technology, the variousorganizations need to publish more and more data, these data includeprivacy more and more, therefore, before the release of the dataneeded to make some data processing, so as not to cause moreprivacy. In recent years, many scholars research on data privacyprotection model are very active, some models have been proposed,such as K–Anonymity model, the L-Diversity model, t-closenessmodel and other data protection model, however, loss of data causedby the data protection model is relatively large. Therefore, this papermade the following work:First of all, summarizes the privacy protection, clusteringalgorithm and rough set theory related to thesis research content.Second, for the t-closeness model is improved, based onclustering is proposed to improve the t-closeness model. The modelmainly for t-closeness model data loss is bigger this defect wasimproved, the first use of density measurement for clustering to reduce the loss of the information in the process of clustering,improved utilization rate of data, effectively protect the privacy. Butin the process of the experiment found that using the clusteringmethod for some character data clustering effect is not obvious, thenthe density measurement is improved, based on rough set K-NearestNeighbor algorithm is proposed. Rough set has unique advantages inprocessing character data, using the rough set for better clusteringeffect.Finally, the above models are protected against single sensitiveattribute, however, in practical applications, sensitive attributes cannot just a single, and in data released there may be more sensitiveattributes. So finally, the paper proposes a t-closeness privacyprotection based on multiple sensitive attributes. The model of thesensitive attribute is divided into two types: The first sensitiveproperties and second sensitive attribute. Respectively, for the twosensitive attribute set different threshold t, making their meetdifferent threshold t-closeness model.