A METHOD FOR PRESERVING PRIVACY IN PUBLISHED MULTI-RELATIONAL SOCIAL NETWORKS

Today, one of most challenging problems in publishing social networks data is preserving privacy of their entities. Until now, some attacks against identity of entities have been known. In particular, neighborhood attacks were defined and a model to preserve privacy of published social networks against 1-neighborhood attacks was proposed. However, this model considered just a simple social network that has only one type of unlabeled relationship between entities. In addition, sensitive relationships that almost exist in real social networks are not considered. In this paper, we consider a more comprehensive and applicable model of social networks including labeled entities and multiple labeled relationships. We also consider sensitive and non-sensitive relationships between entities. In the proposed approach, to anonymize a social network data, we first detect neighborhood graphs and then compute the cost of anonymization. Finally, we will be able to group vertices based on similarity of their neighborhood graphs by using the anonymization cost and anonymize the entire of social network data. We use the k-anonymity approach to develop our new model. Our approach is flexible to help data owners to make proper trade-off between data privacy and data utilization using anonymization cost.