Performance Improvement of Intrusion Detection System for Detecting Attacks on Internet of Things and Edge of Things

Abstract

Edge computing (EC) is a computing methodology that is distributed in nature that brings data storage and computation closer to the place where it is to be used to accelerate response time and the bandwidth save. The Internet of Things (IoT) refers to the collection of all those devices that could connect to the Internet to collect and share data. It is a serious problem to safeguard the IoT environment using a traditional intrusion detection systems (IDSs) due to the diverse types and huge number of IoT devices. The architectural change in Edge of Things (EoT) causes the privacy and security problems to migrate to dissimilar layers of the edge architecture. Therefore, detecting intrusion attacks in a distributed environment as such is problematic. In this situation, an IDS is required. The aim of this paper is to propose an improved IDS models for classification of attacks on IoT and EoT. In order to protect EoT and IoT appliances and devices, an improved IDSs-IoT is proposed by implementing ten different machine learning models. In the first line of this research, normalization technique was performed using the minimum-maximum (min-max) method. Subsequently, dimensionality reduction was performed with principal component analysis (PCA). The light gradient boosting machine, decision tree, gradient boosting machine, k-nearest neighbor, and extreme gradient boosting algorithms were used for classification. We adopt two different cross-validation techniques in this study because it results in skill estimates that have lower bias than other techniques. The experiment was performed on UNSWNB15 dataset. The performance evaluation metrics used are accuracy, area under curve (AUC), recall, precision, F1, kappa and Matthews correlation coefficient (MCC). The findings were bench mark with the state-of-art methods, and our results were superior in terms of accuracy, F1, MCC, and validation dataset.