Knowledge Graph Construction of Network Security Domain Based on Bi-LSTM-GNN

In order to address the increasingly serious threats to cyberspace security, this study focuses on network security events as the research subject. It designs and implements a Knowledge Graph system based on deep learning and proposes a recognition model using Bidirectional Long Short-Term Memory Graph Neural Network (Bi-LSTM-GNN) for constructing the Knowledge Graph in the field of network security. The model effectively captures contextual sequence information through a Bi-LSTM neural network and performs relationship extraction using graph convolutional networks, ensuring the retention of key information while removing irrelevant details. The model achieves high accuracy, recall, and F1 values of 90.2%, 84.5%, and 85.9% respectively on the test dataset, surpassing other models. Experimental results demonstrate the strong performance of the proposed Bi-LSTM-GNN model in identifying knowledge entities in the network security domain.