Research On Anonymous Method Of Effectively Preserving The Community Structure For Social Network Data Publication

Popularity of the Internet makes people’s social model have undergone profound changes, and these convenient platforms also offer great convenience for people’s social contact. With the increasing number of users, the data in network is also growing and becoming an im-portant resource in the Internet. Among them, the analysis of community structure in social network data applications is an important aspect, which can be used to find similar groups, group behavior pattern discovery, in order to develop business strategy recommendation and social behavior studies. However, the data contains a lot of users’ sensitive information, so we can’t publish it in its raw form which will disclose the use’s personal privacy. However, data privacy protection will modify the original data and affect the usefulness of the data publica-tion. Therefore, when the data publishes for analysis, how to get the balance between the pro-tection of individual privacy and the decrease of the data distortion caused by the data privacy is a hot research question in information security and data analysis.The usefulness of data and the purpose of data analysis are closely related, therefore, in order to protect the application value of data, the privacy protection methods for data publica-tion need to consider the purpose of data analysis. However, less of the studies about privacy protection in social network data publication consider specific applications, especially the ap-plication about community structure analysis, and existing anonymous methods usually sacri-fice larger structure information to meet the requirements of anonymity, which has a tremen-dous impact on the analysis of the nature of the community structure, and greatly reduces the practical value. In this paper, the research is carried out by the purpose of community structure analysis for the privacy protection in social network data publication, and proposes the method of effectively preserving the community structure for social network data publication. The main work is as follows:Firstly, we point out technical issues for the community structure analysis in the social network data privacy protection, and then analysis the principles of reducing the data effec-tiveness and the technical limitations when the current privacy protection methods are used in the issue. That is, when the adversary’s background knowledge is any subgroup of the target victim location,k-anonymity is likely to make that the nodes belonging to the same communi-ty are dispersed into different clusters without considering the original community structure, which would blur the boundaries between communities seriously. Besides, most of graph analysis methods usually process atomic nodes and edges, so the^-anonymity graph needs to be reconstructed before analyzing, which is likely to increase the number of edges between communities that are not existing in the original social network, make the connection density increase, and blur the boundaries between communities, in addition, it will bring more uncer-tainty to reconstruct the entire graph, and have a worst effect on the data authenticity.Secondly, for the research questions that how to effectively protect the community struc-ture information of the original social network and the problem of deficiencies of^-anonymity, and through combining the clustering technique with randomly reconstructing technique, we propose a local-perturbation method for data privacy. Our method clusters the nodes with the distance and the original community structure information together as the conditions, so that the nodes belonging to the same community in the original social network could be clustered into one cluster; our method reconstructs clusters one by one, and the connecting structure between clusters remains the same as original social network, that is, the reconstruction opera-tion is controlled in internal community or similar nodes, which greatly reduces the possibility that the density of inter-communities becomes bigger caused by adding or deleting edges, in addition, it also reduces the number of possible graph caused by constructing, thereby in-creases the data effectiveness.Finally, based on the method of local-perturbation data privacy protection, this paper gives a detailed description of the algorithm design and implementation process. Combining the standard used in the community structure analysis commonly with the nature of the graph, we use three evaluation standards to validate the usefulness of the data after anonymization. Ac-cording to the above, this paper gives a detailed system design and implementation process description of each module. Experimental results on three real datasets verify the feasibility of our method and demonstrate that the attacker cannot re-identify the target node with the con-fidence larger than 1/k in published data, and the data usefulness has a greater improvement when the published data is used to the research about community structure.