A K-anonymity Algorithm Based On Jensen-Shannon Divergence

With the rapid development of information technology, data sharing between thevarious organizations are increasingly common.Accompanied by the development of datamining tools/systems, we have to face such an embarrassing situation: data mining systemon the one hand to be able to meet the needs of users to discover valuable information fromthe database, and the other is to limit its mining privacy capacity. Due to the urgent needs ofthe people of Privacy to promote the development of privacy protection technologies.Anonymous method is commonly used in data publishing privacy protection means,since K-anonymity method proposed, its easy to understand, easy to implement and theacclaimed and made many anonymous method for different problems on its basis, such asl-diversity method, m-invariance method, and so on. Although K-anonymity has manyadvantages, but its disadvantages are also obvious. This model for background knowledgeof the attacks, especially the data released continuity can not play a very good protection.This paper main study of a K-anonymity algorithm to prevent privacy in sequentialdata release. First, this paper introduce the study of the status on privacy protection, someconcepts related to it, and K-anonymity model. Next, this paper illustrates some privacyprotection technology used in sequential data release environment may lead to loss ofprivacy, this because of connection between the sequential data, the adversary can mine theconnection to revise his background knowledge and use it to expose the user’s privacy witha great probability.，the normal privacy protection methods with little regard to connectionbetween the sequential data. In order to limit the adversary mining information fromsequential data, in this paper’s algorithm, equivalence classes divide based on theJS-divergence and satisfy K-anonymity principle, the JS-divergence of tuples in the sameequivalence class is less than a given threshold.In order to guarantee the quality of thepublished data, the algorithm draws generalization G.Ghinita proposed, the method usesHilbert filling curve will be multi-dimensional quasi-identifiers are mapped toone-dimensional space and then its the optimal generalization,its improved anonymous datasimultaneously meet K-anonymity and JS-divergence. Finally, by simulation analysis, thealgorithm achieve an acceptable level in preserving privacy and the validity of data.