A level-cut heuristic-based clustering approach for social graph anonymization

Abstract

Privacy preserving data publication is an emerging trend in data publication that focuses on the dual concerns: information privacy and utility. Privacy preservation is essential in social networks as social networks are abundant source of information relating to the behavior of social entities. Social network disseminates its information through social graph. In this paper, we propose a new attack model based on centrality measures. The attack model focuses on identity disclosure problem. Adversary is supplied with centrality measure information of original social graph, which he uses to de-anonymize the published anonymous graph. We have proposed an anonymization model based on level-cut heuristic clustering to generate k-degree anonymous sequence. This step is followed by k-degree closeness anonymous graph construction derived from rich-get-richer phenomenon, transformation, and validation. It is found from the analysis that our proposed approach performs well in assortative networks.