Privacy Preserving Techniques For Incremental Weighted Social Network

With the rapid development and popularity of social networks, more and more people share data and participate in various activities on the Internet. The social network undoubtedly provides broad platform for people’s communication and entertainment, but a lot of social network data which contains personal private information is published to the network. A malicious attacker can make privacy attack using a variety of background knowledge, resulting in leakage of user privacy information. Therefore, how to protect sensitive information in social networks has become a hot topic in the realm of data privacy. In this regard, there have been a variety of social network anonymity techniques in recent years.There are a lot of incremental changes of social networks in real life, that is the nodes and edges of network are increasing as time goes on. Email communication network can be taken as an example. Meanwhile, in real life most of the networks are accompanied with weight information, that is, a lot of social networks can be taken as weighted graphs. Compared with the simple graph, weighted graphs carry more information for the social network, and also will lead to more privacy leakage. Accordingly, this thesis focuses on privacy preserving techniques to support incremental weighted social networks.In this thesis, we first review the existing privacy preserving techniques on social network, and based on this, the social network can be abstracted into a weighted graph incremental sequence and we focus on node identity leakage and edge weight leakage problem in the weighted graph incremental sequence. This thesis first defines Increment Sequence Class Safety Condition, named ISCSC, to guide the anonymous process, and proves that satisfying the ISCSC is a necessary condition to achieve privacy goals. Then with respect to node identity recognition problem, we proposeκ-anonymous privacy protection model based on weight list, to prevent privacy attacking based on the node labels and weight lists in weighted graph incremental sequence, and design κ-anonymous algorithm based on weight list called WLKA, to effectively prevent node identification leakage. In order to further protect the edge weights information and improve the availability of graph data released, we propose κ-anonymous privacy protection model based on hypergraph, to prevent privacy attacking based on the node labels in the weighted graph incremental sequence, and design κ-anonymity algorithm HVKA based on hypergraph, to effectively prevent the node identification and edge weight leakage.Finally, through a lot of tests on real data sets, the test results demonstrate that the WLKA algorithm can effectively prevent the leakage of node identification. The HVKA algorithm can ensure the security of the node identity and edge weight information, and at the same time retain the original structure properties and improve the availability of weight information on the basis of WLKA, but also reduce the time cost of anonymous process.