Intrusion Detection in Marine Networks Based on Feature Dimension Reduction and Graph Convolution

In response to the challenges in the realm of intrusion detection for marine meteorological sensor networks, such as difficulties in model training, inadequate detection performance, and low operational efficiency, we propose an advanced intrusion detection model. This model leverages feature dimensionality reduction, utilizing the Genetic Algorithm based on Random Forest (GARF) technique, to discern the most effective feature subset, thereby streamlining the original network intrusion detection dataset. An Approximate Nearest Neighbor (ANN) algorithm is employed to convert network traffic data into a graph structure, and the Approximate Nearest Neighbor-based Graph Convolutional Neural Network (AGCN) is constructed for traffic classification prediction on this graph-structured data. Our simulation experiments conducted on the NSL-KDD dataset have yielded positive results, demonstrating the model’s enhanced intrusion detection capabilities and efficiency, and affirming its potential to provide robust security measures for marine meteorological sensor networks.