Differential Privacy Protection Method For Dynamic Networks Data Distribution

Information network contains large amounts of data,which is valuable for data mining and data analysis,it also contains a lot of sensitive data simultaneously.On the one hand,a large amount of data is valuable for basic research and cutting-edge exploration.On the other hand,massive data also poses great challenges to users' privacy and security.Because once a privacy leak occurs,it will cause great panic and social instability for the network user group.Data publication for differential privacy has a strict theoretical basis.The real information network is constantly changing according to time,instead of being static.Most existing data publication methods are based on static networks,and less attention to dynamic network data publication.Although the research on dynamic network data publication has increased in recent years,but the network structure of the published network purification map is poorly restored,the data availability is low,and it is not conducive to data mining and analysis in the later.This paper conducts research on dynamic network data publication,the purpose is to improve the usability of data publication,keeping the network structure characteristics of the original graph as much as possible.In order to deal with the problems of low availability and poor network structure restoration of dynamic network data publication,this paper proposes a differential privacy method DP?DNP for dynamic network data publication.This method uses the sliding window to divide the dynamic network into time window according to time node.There are multiple snapshots in a time window,using the method of reservoir sampling to initially filter highsimilarity snapshots,using the graph similarity matching algorithm HED to perform secondary matching on snapshots.Secondly,basing on game theory for dynamic network community discovery,a game theory-oriented dynamic network community discovery method,GCAD,is proposed.The reason for this method is that after the time window division and graph similarity matching,if privacy protection is performed directly,the amount of noise may be too large,so before privacy protection process,in order to reduce the addition of noise,this paper conducted community discovery on the pre-screened snapshots.Using the findings of the community as input,constructing hierarchical random graph.Finally,adding noise to the hierarchical random graph,performing matrix conversion on the noisy hierarchical random graph,converting it to a lower triangular matrix,calculating the average value,and get the lower triangular mean matrix to restore the original network graph and publish it.To verify the effectiveness of the proposed algorithm,we select two real dynamic network datasets in the SNAP graph database to test.Experiments prove that the DP?DNP algorithm can guarantee differential privacy,the reduction of the structural characteristics of the dynamic network is better than the comparison algorithm.This algorithm is feasible and effective.