A Hybrid Deep Learning Framework for Real-Time Network Intrusion Detection in IoT Environments

Abstract

In this study, a hybrid deep learning technique is proposed as the foundation for a real-time network intrusion detection system (i.e., NIDS) for Internet of Things (i.e., IoT) networks. (i.e., NIDS) stands for network intrusion detection system. Learning spatial information is accomplished using convolutional neural networks (i.e., CNNs), while modeling temporal information in network traffic patterns with long short-term memory (i.e., LSTMs) is also used. It is possible to create the architecture in such a way that it is both lightweight and efficient, taking into consideration the limitations of Internet of Things devices, such as the restricted processing power and communication protocols that they must support. Using the NSL-KDD benchmark dataset, the proposed model achieves a detection accuracy of 98.7%, while using the UNSW-NB15 benchmark dataset, it achieves 97.3%, and using the CIC-IDS-2017 benchmark dataset, it achieves 98.1%. Furthermore, the inference time for each packet is less than 2.5 milliseconds. It has been demonstrated through the findings that the CNN-LSTM hybrid model possesses superior detection performance in comparison to conventional machine learning classifiers and single-architecture deep learning models when it comes to identifying both known and novel attacks for various Internet of Things network topologies.