Research On Differential Privacy Protection Method For Heterogeneous Information Network

With the rapid development of social networks,these massive heterogeneous data also pose a threat to privacy leak.In social networks,because of the multiple features of nodes and sides,it can be seen as a complex heterogeneous information network.Accordingly,privacy protection for heterogeneous information networks is urgent to be solved and has social and commercial value.This topic is aimed at studying this problem.The traditional privacy protection method assumes that the network data are independent,but when the data have multiple attributes and strong correlation,the existing privacy protection methods can not play a protective role.The main work of this paper includes: first,we give three kinds of definition of correlation exists in the large-scale heterogeneous network: edges and edges,nodes and nodes,path and path,and then give out the quantitative expression of them.Second,a differential privacy protection scheme based on correlation is designed because the traditional differential privacy mechanism can not process the relevant data.It is proved that the mechanism conforms to the protection condition of differential privacy,and can protect the sensitive information from being leaked.Third,considering the heterogeneous information analysis of complex network structure and information diversity,we used the three indicators of graph data: degree distribution,cut set,the length of the shortest path,based on the three indexes,we study the Laplace noise addition method based on correlation,and introduce sensitivity calculation of correlation coefficient in intervention evaluation the index,to obtain the corresponding Laplace noise.Fourth,comparison and evaluation of the method and the protection mechanism is introduced to the actual network application,through the control of correlation coefficient,not only considers the node and their impact on the global and also consider the node related to other nodes,side impact on the overall situation,in order to global data to protect the effect of the corresponding.By comparing several protective mechanisms,analysis of protective effects of different methods on containing the correlation data of heterogeneous network control,correlation coefficient,privacy control parameters to verify the correlation between the efficiency of differential privacy mechanism.We found that the protective effect of correlation between the subject of differential privacy protection mechanism compared with the traditional method and the interactive method has stronger,to protect sensitive information will not be leaked at the same time the data also has good usability,in general,the correlation coefficient is introduced into the differential privacy role get good effect in heterogeneous network data.