Development of lightweight intrusion model in Industrial Internet of Things using deep learning technique

Abstract

Nowadays, the IDS is being used in conjunction with the IIoT system to reduce the security risk, but on the other hand, the false rate of the IDS is very high. Therefore, in this work, a pre-training method, making use of both a deep neural network and a deep auto-encoder, has been proposed for the quick prediction of assaults with increased accuracy and a reduced false rate. The replicas were expanded using hyperparameter optimization (HPO) techniques. The proposed model delivers an alternative to deep learning construction replicas through an HPO procedure incorporating the Archimedes optimization algorithm. This optimization technique can be used to determine the hyperparameter value and the ideal categorical hyperparameter combination for improved detection performance. The DS2OS dataset is used alongside numerous other indicators to evaluate the efficacy of the developed model. The various existing techniques of assault detection have also been considered to show the effectiveness of the proposed model. Through the comparative evaluation of the outcomes, it is shown that the developed model provides better performance than the other existing models. Eventually, it is discovered that the suggested security paradigm is successful in fending off a variety of internal and external threats.