Preservation of Centrality Measures in Anonymized Social Networks

Social media sites became a pervasive presence in the nowadays society. We can learn a lot of useful information about human behavior and interaction by paying attention to the information and relations of social media users. This information can be public or private. Protecting the private information of the users in social networks is a real concern. Among other perturbation methods, anonymization models have been used to ensure the privacy of social network users. Each anonymization model has its own assumptions about the information that needs protection and various possible attacks attempting information disclosure. A conflicting goal with maintaining the privacy of a network´s information is the preservation of the structural properties of the social network. A good anonymization model would preserve the privacy of social networks´ users and preserve enough information to allow a good analysis of the properties of the social networks. In this paper we investigate how well two anonymization methods preserve the importance of nodes in the network, where node importance is expressed by centrality measures. In the same time, we examine the privacy concerns remnant in these anonymized social networks. Our experiments show that there is an inverse correlation between preserving structural properties of social networks and protecting the privacy of their users. The more information is preserved, the weakest the privacy protection. Also, we learned that different anonymity approaches can preserve information better in certain network types.