A graph modification approach for k-anonymity in social networks using the genetic algorithm

Abstract

Social networks, which have become so popular today, allow their users to share information. The main challenge the users are facing is the security preservation of their information and privacy. Therefore, structural anonymity techniques were introduced that would hide the identity of users. One of the drawbacks of these techniques, which are based on graph modification, is the lack of attention about the structural semantics of graphs. This paper focuses on the popular notion of a privacy protection method called k-degree anonymization and tries to reduce utility loss on the graph. The new k-degree anonymization method, genetic k-degree edge modification, has two steps. The first step includes partitioning of vertices and community detection in the graph. The result of these two determines the needed increase in edges for every vertex in each society to achieve k-degree anonymization. The second step is graph modification using the genetic algorithm by adding some edges between vertices in each community. Average Path Length (APL), Average Clustering Coefficient, and Transitivity (T) are employed to evaluate the method. The proposed algorithm has been tested on four datasets, and the results have shown the average relative performance demonstrates more stability than the other four well-known algorithms. Also, APL criterion in our algorithm is better preserved than all other algorithms; furthermore, Transitivity parameters are the best result in most cases.