Research On De-anonymization Algorithm Of Privacy Discriminate For Social Network Based On Multi-features

Today,social networking services has become a fast-growing business,the development of smart phones has further accelerated the development and application of social networks.Due to the great commercial value of social networks and the huge impact on society.Currently,social networks attract more and more scholars to study.In order to protect the privacy of users,social network owners and service providers often delete or modify personally identifiable information(by adding or deleting edges between nodes to achieve)before the data is published,in order to achieve the anonymity of network data.However,in reality,due to the anonymity of data at the same time need to ensure the availability of data,so that the general structural characteristics of the data is still retained,this simple anonymization of data is vulnerable to various de-anonymization attacks resulting in data leakage,which is data privacy brings new challenges.Anonymity and de-anonymity have always been mutually reinforcing in the area of privacy protection,and they all promote mutual development.Therefore,the analysis and study of the current social network data protection technology flaws and vulnerabilities for social network privacy protection is also extremely important,so the study of de-anonymization algorithm based graph data of social network has great practical significance.The de-anonymization algorithm of the existing social network data mainly adopts the structural features of the nodes as the metrics,that is,according to degree centrality of nodes,closeness centrality of nodes,and betweenness centrality of the nodes as the topological structural features of the nodes,and as nodes a measure of similarity to achieve de-anonymization of social network data.This method has certain shortcomings,because it is just simple to use the topology information of the social network graph,so that the recognition rate of the nodes is not very satisfactory.There are also some other methods of de-anonymyization.For example,a node based a high degree is used to divide the social network data,and then the community detection method is used to implement the node's de-anonymization method.But the problem is that this method of de-anonymization based on community detection increases with the number of nodes in the social network and the complexity of time,the amount of calculation will gradually increase,especially for data sets with nodes above one million,the calculation amount is significantly higher than other de-anonymization algorithms,so this method is not suitable for de-anonymization of large-scale network data sets.To solve the above problems,this paper presents a socialnetwork de-anonymization method based topology structure of nodes and attribute and behavioral characteristics of nodes,that is,by combining the topology structure of nodes within a social network with certain public and behavioral characteristics of nodes to build a metric model to achieve within the social network node de-anonymization.In addition,this paper also proposes a joint topology and behavioral characteristics of attribute inference attacks,this method can be more efficient to infer the attribute information of social network nodes.