Hybrid intelligent intrusion detection system for internet of things

Abstract

The Internet of Things (IoT) has become one of the dynamic and fascinating fields of research with new technology emerging. This can be explained by the ever-rising development of data globally (due to mass deployment of IoT devices) which needs to be protected and secure to avoid misuse. The emergence of IoT and learning algorithms in an area Machine Learning (ML) and Deep Learning (DL) has given new scope to cyber security. The security attacks on IoT networks are possible to detect and prevent intelligently using ML and DL techniques. ML and DL can make traditional attack detection methods efficient, reliable, and robust. The aim of this article is to develop the novel hybrid intrusion (attack) detection model using DL techniques, Convolutional neural network (CNN) and Long short-term memory (LSTM) to achieve better attack detection accuracy. The model is trained and tested using two different datasets namely UNSW-NB1511https://research.unsw.edu.au/projects/unsw-nb15-dataset. and NSL-Botnet22https://research.unsw.edu.au/projects/bot-iot-dataset. to verify the adaptability of the model to different datasets. The developed model has a better detection rate and low false positive rate (FPR) when compared to a state-of-the-art work. The proposed model accuracy is 99.4% with NSL-Botnet dataset and 93% with UNSW-NB15 for binary classification and 92% with NSL-Botnet and 82% with UNSW-NB15.