The Researching Of Link Prediction Algorithm Based On Privacy In Social Networks

With the widespread use of social network, link prediction has become an important research direction. The social network is a mapping from real social relations, there is lots of privacy information that has a great influence to the link prediction. Therefore, it is very important to study the link prediction algorithm of social network using the user's privacy information.In this paper, the user privacy information of social network is analyzed firstly. Then, considering the privacy protection, the link prediction algorithms are studied about the directed social network and directed weight social network. The research work is supported by the National Natural Science Foundation of China(NO. 61401015). The main research contents are as follows:1. Based on the analyses of leakage of privacy information in the social network, the method of differential privacy is used to process user privacy information. The method considered the security of privacy information, and ensured the availability of released privacy information. Theoretical analysis and experimental results show that the privacy information of can be released securely.2. Considering the influence of user privacy information, a model of link prediction algorithm based on user privacy information is proposed. The information about user's privacy and node degree is used to compute the similarity of user information, which makes link prediction methods more conformable to the actual situation. Simulation results show that the proposed model improves the accuracy of prediction results.3. Considering the relationship between the behavior of users in a social network, and combined with the user's privacy information, a link prediction algorithm model based on behavior relations and information of users is proposed. The model analysis user's preference information, and presents a weighted social network link prediction method. Simulation results show that the proposed method can analyze the predicted results quantitatively and has high prediction accuracy.