Researches On Applications Of Differential Privacy Technology Based On Social Network

With the development and popularity of social networks,the security issues of personal privacy data,such as social relationships in social networks,need to be solved.Malicious attackers with relevant background knowledge may collect and analyze specific information randomly or frequently making use of data mining tools,which forces personal social privacy to be exposed,or even be spread rapidly in social networks.Users who experience the convenience of social networks would not expect their private data to be completely transparent.Existing privacy protection methods,such as anonymous technology,are not enough to solve the disclosure of network privacy.This thesis is devoted to use differential privacy to solve the privacy protection problem of the sensitive structural data in social networks.This thesis summarizes the research and development of privacy protection,which mainly focuses on the study of differential privacy protection model and differential privacy protection for social network structure,and also analyzes and compares the advantages and disadvantages of models with different privacy schemes.In terms of differential privacy publication graph and data correlation of social network,following research is carried out.This thesis proposes a differential privacy protection scheme based on community density aggregation and matrix perturbation,and the correlation coefficient k is introduced to preserve the differential privacy and availability of edge related data.First,this scheme adopts the community structure fast identification algorithm to identify the social network structure labels,and the label aggregation of community nodes is realized.Second,it adaptively identifies dense regions of the generated upper triangle adjacency matrix by a data-dependent method and the structure of binary tree.Finally,it reconstructs a noisy adjacency matrix by edge optimal allocation to publish the final network graph,which can further improves the processing efficiency of the algorithm and data availability.Secondly,combining clustering algorithm based on node degree with dK model,a differential privacy processing scheme based on adjacency degree of edge betweenness model and a privacy preserving measurement algorithm based on adjacency degree are proposed.The dK model is used to construct the social network structure,considering the influence of edge betweenness in network;In order to reduce the sensitivity of 2K sequence,an improved model is proposed.The 2K sequence is reordered based on the edge betweenness calculated from the shortest path in every dK tuple,and the whole sequence is clustered into sub-sequences according to the sorting results.In this way the edges of similar influences in the original graph are clustered into the same sub-sequence.Finally,extensive experiments demonstrate that the proposal can achieve more satisfactory privacy protection and data availability compared with the existing classical schemes.