Deep learning for network security: a novel GNN-LSTM-based intrusion detection model

Abstract

The rise in the use of IoT devices in daily life has led to an increase in attacks, making it crucial to protect our devices and information. Intrusion detection system (IDS) is vital in preventing potential attacks. This paper presents a novel IDS architecture using a hybrid GNN-LSTM-based approach. Graph neural network (GNN) is used to extract information from graph-based data, while long short-term memory networks (LSTM) helps learn patterns in the extracted embeddings due to its ability to learn from long-term dependencies in data. We introduce a new mechanism for edge-classification using GNN, eliminating the need for node feature aggregation, followed by edge embedding classification using the LSTM model. We also provide a detailed comparison of our proposed model with state-of-the-art machine learning (ML) and deep learning (DL) algorithms for intrusion detection, demonstrating high accuracy. © © 2025 Inderscience Enterprises Ltd.