AHSWN Home • Issue Contents • Forthcoming Papers
Deep Learning Enabled Privacy Preserving Techniques for Intrusion Detection Systems in the Industrial Internet of Things
Radha D and M.G Kavitha
Industrial Internet of Things (IIoT) becomes a hot research topic which makes use of sensors and actuators with computing and communication abilities to transform the way of data collection, communication, and processing. The rising ubiquitous capability results in innovative Industry 4.0 (also referred to as Industrial Internet) applications for enhanced production and effectiveness in several industrial sectors namely mining, energy, healthcare, agriculture, and transportation. Though IIoT offers significant advantages in several applications, security and privacy are considered as the challenging issues that exist in the design of IIoT, which can be resolved by the use of using intrusion detection systems (IDS). The recent advancements of machine learning (ML) and deep learning (DL) algorithms pave the way to design effective IDS techniques for IIoT environment. In this view, this paper presents deep learning enabled privacy preserving technique for IDS (DLPPT-IDS) in IIoT environment. The proposed DLPPT-IDS technique aims for determining the existence of intruders in the IIoT networks by the use of DL and feature selection techniques. In addition, the proposed model involves the design of improved salp swarm optimization based feature selection (ISSO-FS) technique which integrates the concept of hill climbing with the conventional SSO algorithm to improve the global optimization capabilities, shows the novelty of the work. The proposed model also uses bird swarm algorithm (BSA) with stacked sparse autoencoder (SSAE), named BSA-SSAE classifier for the detection and classification of intrusions in the IIoT environment. The performance of the presented technique can be examined using benchmark dataset and the experiments results showcased the promising outcomes of the proposed model over the other recent state of art IDS techniques with the precision of 99.16%, recall of 99.33%, F-measure of 99.28%, and accuracy of 99.49%.
Keywords: Intrusion detection system, Industrial Internet of Things, Deep Learning, Metaheuristics, Feature selection