Intrusion Detection Systems for IoT Attack Detection and Identification Using Intelligent Techniques

Abstract

The Internet of Things (IoT) and its connected objects have resource limitations, which lead to weak security concerns over the IoT infrastructures. Therefore, the IoT networks should always be attached with security solutions. One of the promising security solutions is intrusion detection system (IDS). Machine Learning (ML) algorithms become one of the most significant techniques for building an intelligent IDS based model for attack classification and/or identification. To keep the validation of the ML based IDS, it is essential to train the utilized ML algorithms with a dataset that cover most recent behaviors of IoT based attacks. This work employed an up-to-date dataset known as IoT23, which contains most recent network flows of the IoT objects as benign and other flows as attacks. This work utilized different data preprocessing theories such data cleansing, data coding, and SMOT theory for imbalanced data, and investigating their impact on the accuracy rate. The study's findings show that the intelligent IDS can effectively detect attacks using binary classification and identify attacks using multiclass classification.