AM-DRCN: Adaptive Migrating Bird Optimization-based Drift-Enabled Convolutional Neural Network for Threat Detection in Internet of Things

The Internet of Things (IoT) is a network that connects devices and sensors which enables the interconnection of people to everything and everywhere. The high network traffic has led to various cyber attacks, so efficient detection is needed. However, the existing methods pose drawbacks such as high complexity, latency, and computational issues. To overcome the limitations, an Adaptive Migrating bird optimization-based Drift-enabled convolutional neural network (AM-DrCN) classifier is proposed for effective threat detection. The Adaptive Migrating bird optimization (AMO) algorithm is employed to tune the hyperparameters of the model. In this research, the K Nearest neighbor (KNN) technique performed missing data imputation for estimating the missing details of unevenly distributed data, which enhances the overall efficiency of the model in threat recognition. Furthermore, the framework achieved improvements in the precision of 94.93%, accuracy of 94.75%, F1 score of 94.75%, and recall of 94.57% at 90 TP.

I. Introduction

Over the past period, the ecology of human life, business methods as well as production have changed resulting in an Industrial 4.0. The four important elements required for Industrial 4.0 are IoT, Artificial intelligence (AI), Cloud, and Robotic Process Automation (RPA) [1] [2]. The IoT model is generally comprised of a three-stack framework comprised of network, sensing, and implementation layers [3]. The advancements in the technologies have occurred considering the features of IoT such as perceiving, interconnecting, processing, and implementing information at various levels [4]. To optimize the industrial process control of IoT, the rise of the Distributed Control System (DCS) gives more computerization by utilizing cloud computing. Industries are massively adopting IoT prompting automation for improving performance with error-free surveillance of data as well as reducing the privacy issues happening in diverse enterprise networks [5] [6]. Personal privacy, confidential information, and essential information have attracted hackers due to the rapid increase of traffic in the system. The extremity and intensity of worldwide network intrusions are also widely increasing. According to the World Health Organization (WHO), the attacks occurring have reached 500 % annually [7].