Research On Privacy Protection Algorithm Based On K-anonymity

With the advent of the big data era and the rapid development of Internet technology, more and more information have been published or shared online. Internet provides a platform for the current scientific research, information exchange, data sharing, but it also put forward new challenges for privacy protection technology. On the one hand: A lot of research institutions want to get low-cost information resources needed for data analysis; on the other hand: The owner of the released information do not want their private information has been compromised. Thus, the private information data published technology has become a new research direction in the field of information security. k-anonymous model was an effective way to protect the privacy data. The idea of k-anonymous model is simple, easy to implement. So, it is the current mainstream model of privacy protection. However, there have its own limitations. Through studying the anonymity model deeply, this paper pointing out the lack of k-anonymity model, and for the k- anonymity defects proposed a new improved model. The main contributions of this thesis can be summarized as follows:Firstly, based on previous research and solutions on the k- anonymity, studied the effect of k-anonymity privacy, existing problems, and existing improve algorithms for k-anonymity at home and abroad, and analyzes several more influential improved algorithm.Secondly, this paper analyze existing anonymous algorithms based on the clustering. In order to better reduce the amount of information loss, this paper proposes a new anonymity algorithm based on clustering--Max DD algorithm.Combined the clustering algorithm based on maximum dissimilarity with(a, k)-anonymity algorithm, and improved generalization lattice of the classical models.Thus achieve better privacy protection while minimizing the effect of the amount of information loss.Thirdly, through research on the micro-aggregation k-anonymity algorithm, on the basis of the largest clusters of dissimilarity above in the last chapter, in order to achieve anonymity processing, Micro-aggregation technology instead of the traditional generalization techniques. Combining this technique with L-diversity algorithm, it can reduce the information loss and improve the practicability of data.Fourthly, in this paper, through the experiment, comparing the two improved algorithms with existing algorithms on the time efficiency and information loss, and experimental analysis is presented.The proposed algorithm by this paper, was used commonly for mixed data in micro data sets.