Research Of Link Prediction Method Based On Differential Privacy

Link prediction is one of the research hotspots of graph models and network data analysis.This method predicts the possibility of connecting edges between pairs of nodes in a complex network that have not yet formed edges by analyzing known network nodes and related information such as their structure.However,the improper use of private information such as nodes(user attributes)and connections(user relationships)used in the link prediction process has attracted people’s attention.The private data contained in the real network may lead to the disclosure of the real user information,It will bring negative impacts and even economic losses to users and information publishers,which will seriously affect the prediction accuracy of the link model.Based on this,this paper focuses on the privacy protection of network link prediction and conducts research,which mainly includes the following two parts:(1)Propose a random disturbance link prediction algorithm with privacy protection ability.It is proved that the random disturbance mechanism satisfies the -differential privacy.By adding random noise to the original network data,simple random disturbances based on the edge differential privacy are respectively proposed.Mechanisms and link prediction algorithms that optimize random disturbance mechanisms.Finally,experiments with real network data show that the proposed method has good privacy protection capabilities and link prediction capabilities.(2)Propose a network link prediction model based on the differential privacy Laplace mechanism and solve the algorithm.Consider deleting the network edges and nodes to calculate their corresponding global sensitivity,and add Laplace noise to the query function to realize the privacy protection of user data Finally,the availability and protection of the Laplace mechanism network link prediction algorithm is verified through real network data.