Research On Privacy Preservation In Social Network

With the improvement of data sharing and data mining technology, people get more information, and meanwhile, the disclosure of personal data privacy also abtains more and more attention.Privacy preservation mainly consists of two aspects:the preservation of sensitive knowledge and the preservation of sensitive data. Sensitive knowledge mainly refers to the association rules, classification rules extracted from the database and so on; sensitive data is the data that can be mapped to the individual, resulting in the privacy of individual data disclosed. With the development of Internet technology, social network applications are becoming increasingly popular, and also make privacy data not only limited to relational tables, but also expanded to a large number of complex network data. This paper mainly focuses on social network data privacy preservation. Following works have been proposed:1) Summarize the privacy preservation technology of relational database, including attacking models, anonymization models, anonymization methods, information utility and background knowledge.2) Introduce the social networking related concepts, composition, and main research fields. Summarize the privacy preservation technology of social networ, including privacydefinitions, anonymization models and anonymization methods.3) Propose (k, u)-PN anonymity of path nodes in social network.(k, u)-PN anonymity limits the total number of nodes on the paths and the proportion of paths that each node belongs to. It avoids the attack when attackers know little about the target node, but have some knowledge of the neighbor nodes, so as to identify the target node by the paths between its peripheral nodes. PN-anonymity algorithm is proposed by adding edges/nodes, so that the paths become diverse and the total number of nodes in the path increases.4) Propose k-sum_value anonymity. Existing anonymity model applies to unweighted graph. This article will extend the concept of node degree in unweighted graph to weighted graph. The degree of a node is the sum of edge weights whose one endpoint is this node, to distinguish the degree in unweighted graph, denoted sum_value. k-sum_value anonymity model is proposed, which limits the number of nodes with the same sum_value. An algorithm is proposed to make the graph satisfy k-sum_value anonymity by adding weights, edges and splitting nodes.