A Novel Deep Learning-based Approach to encounter cyber threats in IIoT

Abstract

Internet of Things (IoT) is a growing field and it has reached the multi-million dollar market. The research in the field of IoT, networks and AI is in initial stages due to the growing nature of IoT market. The IoT devices are used in many different applications to automate processes. In Industrial Internet of Things (IIoT), millions of such tiny devices are used to automate the quality assurance, safety protocols and other industrial processes. Due to resource constraint nature of such tiny devices, security is a big challenge for researchers to detect the security-based threats in IoT. Hence, intrusion detection is a big problem in IoT. In this paper, a novel approach to detect intrusions and cyber threats is proposed. In the proposed approach, the out class deep learning based algorithms are used to detect cyber threats in IoT. For this purpose, a latest data set named as Kitsune is used. This data set is already pre-processed and contains rich feature sets. Moreover, it has latest data of 9 types of attacks. In the proposed strategy, work has been done on a single type of attack namely Mirai Botnet and four different algorithms LSTM, GRU, DNN, RNN with the combinations of CNN1d, CNN2d and CNN3d are used. The simulation results show that the proposed approach with an accuracy of 99.73 outperforms traditional approaches.