Research For Protecting Privacy Of Social Network Data Based On Relevance Degree Perception

With the development of the Internet,social software has become an indispensable part of people's lives.People can use social networking sites or social software for ordinary daily communication,academic communication,voting and other activities.The massive growth of online social networks has created unprecedented unique opportunities for social network analysis.People can analyze these data,and the analysis results can have a lot of commercial value and social value for researchers.For example,data managers can use relevant data for sociological research,infectious disease research,business model analysis,etc.The value of social network data is gradually being reflected.In order to realize the value of social network-based data analysis and application,it is necessary to ensure data security before publishing social network data.Therefore,effective privacy protection analysis technology for such data is crucial.The current deficiencies in the protection of social network data privacy include that existing work is to modify the nodes and edges in the graph,but such graph and edge modifications are easy to destroy the structural information and correlation information of the entire graph.Social network data is a complex structured data.Not only the structure is related to the edge,but the attributes of the nodes are also related to the edge.If you do not consider the relevance of the social network data,only the privacy protection of the sensitive information of the edges or nodes,The correlation of the original data is greatly damaged,which will lead to a decrease in the utility of the data.Secondly,in existing privacy protection work that considers attributed social network data,consider that the correlation in the data is relatively simple,but this is not in line with the actual situation.The attributes of nodes are multi-sensitive attributes,including sensitive and non-sensitive attributes,multiple sensitive Attributes are related to each other and have complex inlining.These attributes have multiple associations.The core problem solved in this article is to ensure the privacy information of the social network data edges and attributes,and at the same time to ensure the effectiveness of the social network data.In order to solve the above shortcomings,the innovation and main work of this article include the following aspects:(1)Aiming at the privacy protection problem in the published graphs of social networks with attributes,the definitions of structural association degree,attribute association degree andcomprehensive association degree between nodes are proposed.The comprehensive association degree is used to cluster the original graph to balance the nodes.The closeness of the connection and the closeness of the attributes are beneficial to protect the correlation between the structure and the attribute during the process of noise addition,reduce the information loss value of the structure and the attribute,and then improve the utility of the social network data.(2)In each subgraph,we use the hierarchical random graph model to represent the structure of each subgraph,and use the Laplacian mechanism on the hierarchical random graph model to add noise to the connection probability in the sample tree to meet the differential privacy protection method.Realize the privacy protection of the social network data's side information,instead of adopting the direct side-by-side random perturbation method,thus improving the structural utility of the social network data.(3)In the work of protecting privacy for multi-sensitive attributes,we distinguish between sensitive attributes and non-sensitive attributes,which are regarded as quasi-identifiers and sensitive attributes,respectively.Then,according to the characteristics of different privacy protection methods,a variety of privacy protection model fusion methods are used.Based on the quasi-identifier,the generalization tree rules are designed to satisfy the k-anonymity method,and the generalization tree rules are designed based on sensitive attributes to satisfy the l-variety.Sex.The generalization tree is a different generalization tree customized in advance according to the meaning of different attributes.(4)Finally,experiments were conducted on three real data sets,and the experimental results verified that the method proposed in this paper ensured the same privacy protection strength and better protected the structure and attribute information of social network data.Improve the consistency and utility of the structure and attributes of the image to be published and the original image.