| In recent years,with the rapid development of internet technology,social network software has became more and more popular,such as micro-blog,WeChat and Facebook,the social network application provides a convenient platform for people to communicate,but also produces a large amount of information about the user.This information has wide application,such as advertising,commodity recommendation and social behavior prediction.Social network data contains a large number of sensitive information,including personal attributes(such as career,salary,etc.),personal behavior information(such as personal social relationships,etc.),if the information is distributed without processing,it may infringe on the privacy of the users.Therefore,the issue of privacy protection of social network data publishing becomes the focus of many researchers.Current work focuses on static network analysis,however,many applications involve the development and change of the network.Different from static network,the dynamic network data privacy protection is put forward higher request.It needs to ensure that the data can meet the requirements of anonymity at a certain time,moreover,due to the inherent relationship between the data at different times,the attacker can get more privacy information by comparing the data published in time,but also to ensure the security of privacy information,so,the existing privacy preserving methods for static social network analysis are not suitable for the need of privacy protection.In order to solve the above problems,this paper proposes a privacy preserving model to resist label-neighborhood attacks in dynamic releases of social network.First of all,through the analysis of the current status of privacy protection in dynamic releases of social network,this paper points out the problem of privacy leakage in dynamic releases of social network,and analyze the existing privacy protection methods used to solve the problem of social network privacy issues:the existing methods of grouping individuals with sensitive labels are based on the similarity of neighbor labels,without considering the neighborhood structure information in social networks.In the process of obfuscation of the label-neighborhood graph with sensitive labels,this operation makes the social network graph adds a lot of noise edges,impacts the availability of network data;although each release of the social network can meet the dynamic needs of network privacy protection,the attacker can still be based on the background knowledge of the social network graph to be published for many years,it is possible to identify only the individual with sensitive information.In addition,if the whole graph is treated with an obfuscation method,this operation against analysis of data,many individuals who do not need to be protected are also processed by randomization,which causes the loss of the original social network structure information and reduces the availability of data.Secondly,this paper proposes a privacy preserving model for dynamic-l-diversity,which is based on the security issue of social network.By using this method:1.In view of the existing grouping method only consider the label-neighborhood information,did not consider the problem of neighborhood structure information nodes in the original social network,this paper puts forward the method of grouping with sensitive labels according to individual structural similarity.2.Aiming at the problem that we can uniquely identify the individuals with sensitive label by comparing the label neighborhood information in the dynamic releases,this paper proposes an anonymous method to satisfy the l-diversity in the dynamic social network.The random perturbation method is used to change the structure of the graph,the edges are added randomly,and then the method is used to randomize the neighbor graph with the sensitive label.The obfuscation method of social network graph,makes each edges have the corresponding probability exists in the social network,make the attacker can not uniquely identify the individual with sensitive labels with probability higher than 1/LFinally,based on the dynamic network,the method of protecting the privacy of the label-neighborhood is proposed,the design scheme of the system is presented in detail.Three evaluation indexes for the performance analysis of social network are designed,which are average degree,clustering coefficient and graph structure information.The evaluation results show that the privacy protection method in dynamic release of social network can preserve the attributes of the network structure,while ensuring the privacy of individual information. |
