Enhancing IoT intrusion detection system with modified E-GraphSAGE: a graph neural network approach

In network intrusion detection, graph neural networks (GNNs) have gained remarkable attention in addressing cybersecurity threats. This research addresses the growing cybersecurity challenges by introducing the E-GraphSAGE model, which leverages GNNs to enhance network intrusion detection. Unlike previous approaches, this model uses a modified message function which computes messages by concatenating the source node and edge features and passing them through a linear transformation. This enables the model to take both features of the source node and edge during the message computation to provide flexibility in modeling complex relationships between nodes and edges.