An Uncertain Graph Privacy Preserving Scheme Based on Node Similarity in Social Networks

Abstract

Since the rise of social networking sites, social networking has become an integral part of daily life, containing not only a large amount of worthwhile data for research, but also personal information about individuals. In the analysis of social network data, users' private information may be revealed, so ensuring user privacy before data publication is important. Focusing on this problem, we propose an Uncertain Graph scheme based on Node Similarity (UG-NS), which can not only preserve user privacy in social networks but also maintain high data utility. In our scheme, we obtain the uncertain graph in two steps. Firstly, we assign edge probabilities to the edges of the original graph based on the node similarity. Secondly, we add some noisy edges and add Differential Privacy noise to these edges. In contrast to other existing uncertain graph schemes, our scheme considers the structural characteristics of the original network when assigning edge probabilities, using the Node2Vec model to calculate the similarity between node pairs. In addition, our scheme adds noise to the graph, which can provide better privacy preserving. A number of experiments have shown that our scheme provides a well-balanced between user privacy and data availability when transforming certain graphs into uncertain graphs.