Affiliations: [a] CICAEET, Jiangsu Engineering Center of Network Monitoring, Nanjing University of Information Science & Technology, Nanjing 210044, Jiangsu, China | [b] School of Computer & Software, Nanjing University of Information Science & Technology, Jiangsu, Nanjing 210044, Jiangsu, China | [c] School of Economics & Management, Nanjing University of Information Science & Technology, Nanjing 210044, Jiangsu, China
Correspondence:
[*]
Corresponding author: Tinghuai Ma, CICAEET, Jiangsu Engineering Center of Network Monitoring, Nanjing University of Information Science & Technology, Nanjing 210044, China. E-mail: thma@nuist.edu.cn.
Abstract: The increasing population of online communication and telecommunication has interested scholars and researchers considering their social networks. These social networks datasets play an exceptionally important role in the research of data mining. However, large amounts of social network data are produced by using social networking applications. And these data inevitably contain a large amount of personal privacy information. Therefore, in order to avoid disclosure of privacy, the data holders need adopt privacy protection before these data are released. Furthermore, most current methods of privacy protection are based on the simple graph only. The weight values on the edges represent the tightness between the nodes. The algorithm based on weights in privacy protection field is still relatively rare. In real social networks, the weight can indicate tightness between two individuals of social relations. The weight may be as attackers’ background knowledge to re-identify the target individual and lead to loss of privacy. In this paper, we consider protecting the weighted social networks from weight-based attacks and propose a method based on the weighted social networks, named k-weighted generalization anonymity (KWGA). And This method combines k-anonymous with generalization method to ensure the security of the social network data when it is published. In order to ensure the higher validity of privacy protection, this paper introduces a concept of the weight difference to reduce the modification for weight graph. Finally, we firstly use real dataset C-DBLP to verify the validity of our method perform much better than the Fast k-degree anonymity (FKDA) in average clustering coefficient (ACC), global clustering coefficient (GCC), average path length (APL) and rate of edges change four aspects. Furthermore, we also use two common datasets in the research of weighted graphs to verify our algorithm.
