Research On De-anonymization Method For Social Network Data Based On Structured Feature

With the widespread use of the Internet,social network has become an important carrier of information dissemination and sharing in contemporary society.For the sake of commercial interests and research,social network privacy protection and privacy attack have become an important research topic in recent years.In order to protect the user's privacy,privacy protection technology must be performed before releasing the original data.The anonymization method is one of the important methods in the privacy protection technology,and its performance depends on the exploration of deanonymization methods.Starting from privacy attack research,this thesis focuses on de-anonymization of anonymous social network data and mobile trajectory data.The main work and contributions are as follows:(1)Aiming at de-anonymizing social network data,this thesis proposes a deanonymization method based on measurement of similarity of nodes that integrates its structural features and attribute features.This method analyzes and utilizes the similarity differences between the structural features of the social network graph and the node attribute features,and performs similarity matching to achieve the mapping between user nodes so as to achieve the de-anonymization of the anonymous users.When implementing the mapping between nodes,in order to reduce the mapping space and complexity,this thesis selects the node with high node degree as the initial mapping node and adopts breadth-first search method to traverse all nodes.The experimental results verify the feasibility and effectiveness of the de-anonymization method proposed in this thesis.In addition,after completing a certain mapping,the associated attributes of the anonymous node are connected to the node of auxiliary dataset graph to realize the correlation estimation of other private information.(2)Aiming at the problem of de-anonymization of mobile trajectory data,on the basis of the analysis of characteristics of the moving trajectory,this thesis proposes a de-anonymization method based on the trajectory space-time feature by using the frequency difference of the key road segments in the trajectory,so as to improve the accuracy of de-anonymization of the trajectory data.Firstly,this thesis uses TF-IDF to calculate the track segment frequency of the trajectory to construct eigenvectors.Then dimension of the constructed eigenvectors are reduced by the method of Principal Component Analysis.Finally,the similarity matching between trajectories is performed on feature vectors of reduced dimension,so as to identify the users corresponding to the anonymous trajectories.The experimental results show that our method outperforms other de-anonymization methods in accuracy,which verifies the effectiveness of the proposed de-anonymity method.The research in this thesis provides some references significance for privacy attack and protection in social network.