Heterogeneous GNN with Express Edges for Intrusion Detection in Cyber-Physical Systems

With the ever-increasing population of Internet-of-Things (IoT) and Cyber-Physical systems (CPS), cyber attacks can result in significantly more severe consequences. In this paper, we introduce a novel data modeling technique using a heterogeneous graph with Express Edges to enhance the attack detection capabilities of machine learning models. Addition-ally, we present the first-of-its-kind performance benchmark of representative heterogeneous graph neural network (HGNN) algorithm variants using multiple network intrusion detection system (NIDS) datasets. Our primary aim is to assist CPS defenders in achieving optimal attack detection efficacy against cyber threats.