K-anonymity Privacy Protection Based On Fast Search And Find Of Density Peak

With the development of information technology,data-rich resources to bring convenience to people's lives at the same time,information security is also facing huge challenges.Some social networks before the data is published in order to prevent the disclosure of privacy information,privacy protection for these data is required,however,the traditional methods of privacy difficult to balance privacy and security,and data availability.Therefore,to ensure data privacy and security while improving the availability of data,become a research topic in recent years.Present in a large number of privacy protection methods,k-anonymity is a common and effective method.In the context of big data today,k-anonymity method based on clustering is one of the recent research focus in this area,compared with the traditional k-anonymity algorithm,by clustering the data operation,increased data availability.But there are still some problems,such as accuracy is dependent on the initial set of these information leading to more losses,poor adaptability for different types of data sets.In response to these issues,this paper presents an improved algorithm,work as follows:(1)Based on fast search and peak density clustering algorithm for k-anonymity model for improving the proposed anonymity algorithm based on CFSFDP(K-CFSFDP).Dividing the data,the algorithm can automatically identify in the course of cluster centers and arbitrary class can be found improved classification accuracy,reducing the loss of the anonymous information and running time of the algorithm.(2)CFSFDP based on semi-supervised algorithm is proposed.By introducing a supervisory information(in pairs),to guide the clustering process,improve the accuracy of CFSFDP algorithm for clustering.(3)K-CFSFDP anonymous CFSFDP algorithm based on semi-supervised algorithm was improved,K-CFSFDP privacy protection based on semi-supervised algorithm is proposed to further reduce information loss.Through Adult in the UCI machine learning database simulated data sets,proof of the effectiveness of the algorithm.