Privacy Protection For Weighted Social Networks Based On MST-clustering Sub-graph Generalization

Along with the rapid development of Internet technology,social networking sites?SNS?are increasingly emerging.As information is easily exposed to the network,social network privacy protection becomes focus of study.Classical technologies for privacy protection are usually used for relational database or tables,which including the generalization technology,anonymity technology and data perturbation technology are mature.The anonymity technology is complex but with high data utilization.The latter two ones are simple and of good privacy protection but with low data utilization.However,classical technologies can not be directly applied to social networks,but their basic ideas also inspire studies.With considering both protection privacy and data utilization,based on the MST?minimum spanning tree?clustering and sub-graph generalization,a privacy protection method is proposed for weighted social networks.Main contributions of the study are?1?For weighted social networks,presents one privacy protection approach based on MST clustering sub-graph generalization.Firstly,to determine the MST of one given weighted social network by removing loop edges.Then to remove the unusual edges called“inconsistent edges”.Finally,to recognize clusters?sub-graphs?that are connected components of the MST;?2?Proposed three sub-graph generalization approaches for privacy protection of weighted social networks on the basis of MST clustering.The three sub-graph generalization approaches are called the Avg Wedge,Max Wedg and Min Wedg sub-graph generalizations.In the Avg Wedge sub-graph generalization approach,edge values of each sub-graph?cluster?are generalized/re-valued as the average edge value of the sub-graph?cluster?.And the similar operations for the other two sub-graph approaches.At the same time,the so-called "inconsistent edges" or loop edges are treated in one reasonable way.?3?Experiments and result analysis indicate workability of the proposed method.The proposed MST clustering-based sub-graph generation privacy protection algorithm for weighted social Networks?called MST-SGG?is applied to experiments.The experimental datasets include the Pattern Recognition example dataset?PR?and several datasets?i.e.Shr.net,write.net,gr3₅3.net?in the Pajek test datasets?"Test Networks"?.Experiments and result analysis indicate that the proposed privacy protection algorithm works well.Conclusion: To improve privacy protection of the weighted social network,in general,it will inevitably decline its data utilization.Then study of privacy protection aims perfecting its balance.Since the boundaries?the so-called "inconsistent edge" and loop edges?of different clusters?sub-graphs?are remained in the proposed MST-SGG generalization algorithm,data utilization of the proposed algorithm is high especially for data mining while it can also provide good privacy protection by the sub-graph generation operations.