An AK-Secure Privacy Preserving Approach Based On Automorphism In Social Networks

The rapid development and popularity of the online social networks (OSN) has gained tremendous popularity and attracted a growing number of researchers and developers to engage in scientific research. However, OSNs store a large amount of valuable personal information, the concern for privacy leakage hinders the data exchanging and sharing in OSNs. Therefore, how to protect the privacy information contained in OSNs during the releasing of social network data has became a hot issue in the state of the art techniques.In the area of privacy preserving in OSNs, the research has been studied extensively. The existing privacy preserving technologies in social networks are mainly based on the thought of K-anonymity and a small part of the thought of data disturbance. At present, the privacy preserving models based on graph isomorphism have the advantages of withstanding diverse attacks and providing strong protection. Those works drew a lot of attentions, but these privacy preserving approaches based on the models still can not effectively prevent edge leakage and path length leakage in social networks.In this paper, we design a secure and high utility privacy preserving model, called AK-Secure, to prevent node identity attack, edge leakage, and path length leakage effectively. The model guarantees the probability of any node in the model being identified is less than or equal to1/K by continuing the thought of K-automorphism, and guarantees the probability of any edge in the model being identified is less than or equal to1/K by adding constraints on edges and paths between nodes, so there is no edge leakage and path length leakage in the model. On the basis of the AK-Secure privacy preserving model, we propose a graph anonymous algorithm, which constructs an anonymous graph satisfying the AK-Secure privacy preserving model by minimizing information loss, preventing privacy leakage and guaranteeing high data utility of the released anonymous graph, which is practical significance for the releasing and sharing of social network data. At the same time, for the privacy leakage of node attribute information in the data releasing of social networks, we propose a secure rule of releasing node attribute information in actual social networks, which can prevent node privacy leakage effectively. Extensive experiments on real data sets show the correctness, effectiveness of our AK-Secure graph anonymous algorithm, and the high data utility of the released anonymous graph.