FIR-GNN: A Graph Neural Network Using Flow Interaction Relationships for Intrusion Detection of Consumer Electronics in Smart Home Network

In the smart home scenario, the Consumer Internet of Things (CIoT) deeply integrates into daily life with various Consumer Electronics (CEs) like home cameras, smart speakers, smoke/fire detectors, VR/AR game boxes/handles, and future home medical terminals. However, CEs face multiple risks due to attack concealment and protocol differences. Against this backdrop, embedding Network Intrusion Detection System (NIDS) in the smart home gateway is proposed. Despite Machine Learning (ML) and Deep Learning (DL) enhancing network intrusion detection, challenges remain in sample collection, traffic feature expression, and gateway resource constraints. To address these, we propose FIR-GNN. It constructs a FIRG graph for traffic pattern capture, uses edge-wise graph attention in FIR-GNN for semi-supervised learning, and selects features by SHAP to cut resource consumption. Experiments show FIR-GNN improves classification performance by 3-5% on BoT-IoT and CICIDS2017 data, safeguarding smart home CEs.