Rearch On Anonymous Privacy Protection Algorithm Based On Clustering

With the continuous development of information technology networks and applications, a large amount of data have been accumulated. People could get valuable information by data mining from the clutter of data. Analysis of massive data produces enormous economic benefits and facilitates people's lives.The direct release of data, at the same time, is inevitably brought privacy disclosure problems. Anonymity is one of the main technologies to achieve privacy protection in data release. This thesis carries out a study of anonymity privacy protection based clustering algorithm.In allusion to the existing problems that the k -anonymous algorithm based on clustering was more seensitive to isolated points, more information loss, and less efficient implementation,this thesis proposes an anonymous improved algorithm based on clustering optimization. The algorithm eliminates the sensitivity of the isolated points of the clustering results by dividing the isolated points, in order not to make its involve in the whole process of clustering; what reduces the number of iterations in the process of clustering through calculating the initial clustering center. Clustering convergence is so fast and so easy to obtain more accurate clustering results that the information losses have been reduced in the subsequent generalization process, that efficiency of k -anonymous privacy protection algorithm based on clustering has been improved effectively. In this thesis, the k-anonymous improved algorithm based on clustering is analyzed from the aspects of security, effectiveness and complexity of the algorithn, and the algorithnm is experimentally tested from the aspects of information loss and execution efficiency.Aiming at the existing problems that the l-diversity anonymous privacy protection algorithm based on clusrtering is more information loss, is susceptible to skew attacks and similarity attacks, this essay proposes a l-diversity anonymous improved algorithm based on sensitivity constraint. The algorithm achieves better clustering results by selecting preferentially high density points in the clustering process. And equivalent sensitive group frequency constraints have been set according to the sensitivity of attribute values. The frequency of sensitive attribute values and equivalent sensitive group frequencies in the same equivalence class are constrained simultaneously. What the above operated enhanced data security release and reduced information loss. Eventually in this thesis, -diversity anonymous algorithm is given the results, and algorithm attack vulnerability and loss of information as well as the algorithm efficiency of the algorithms are analyzed detailedly.