Preserving Privacy for Hubs and Links in Social Networks

Online social networks provide valuable information for researchers. However, users' privacy will be exposed when the social network data is published. Social relationships are connected with our social identification which represent users' sensitive message. In addition, there are some nodes in the social network that are closely connected to others, called hub nodes. These hub nodes play an important role in the social network. Many proposed methods only protect the privacy of social relationships, the hub nodes with a large amount of neighbors are exposed to the adversary. Therefore the adversary can easily find out the identity of these nodes, and then inferring the probability of the existence of a link, although the link is perturbed. To tackle this problem, we proposed a random walk based algorithm which can protect both hub nodes and social relationships. Through a random degree perturbation algorithm, the performance of link privacy preserving method can be improved. We conducted comprehensive experiments over the real-world social network dataset to evaluate the anonymity level and utility of the transformed graph. The experimental results show that the proposed mechanism can achieve higher link privacy and has better utility, outperforming the state of the art approaches.