GNN_NEI: A Network Entity Identification Model Based on Graph Neural Network

Network Entity Identification (NEI) is an essential issue in network security, which involves identifying various classes of entities with IP attributes on the Internet. Traditional NEI methods tend to have the disadvantage of being ineffective in the face of large-scale, complex network data; these GNN-based heterogeneous models require all nodes to have attributes. However, this is only sometimes satisfied in a network topology graph. In network topology graphs, specific entity node attributes are typically said to be missing, which leads to a decline in the accuracy of NEI task. The missing attributes can be obtained learnable by utilizing network entity features combined with graph neural networks embedded in the nodes of the graph to get a numerical vector representation of the entity nodes. And propose a graph neural network-based NEI model (GNN_NEI), including data processing, node information update and graph neural network classifier. GNN_NEI model performs an attribute aggregation update step by calculating the weights between nodes, followed by classification, and finally outputs the entity types. The GNN_NEI model has significantly improved all the indexes compared with other methods, and the identification accuracy is over 92.37%, and the highest can reach 95.42%.