A comparative study of supervised Machine Learning classifiers for Intrusion Detection in Internet of Things

Abstract

In recent years, the field of computer networks and the internet has experienced exponential growth. At the same time, it also raises issues with respect to security. The standard responsive techniques, such as antivirus, firewalls, spyware, and authentication mechanisms, provide security in many areas but still face the challenge of intrusion attacks and viruses. Meanwhile addressing the limitation of intrusion attacks and viruses, an Intrusion Detection System (IDS) has already been proposed by earlier researchers using some Machine Learning (ML) Classifier technique but it also has some major drawbacks like dealing with old datasets, less number of attack classes, unable to monitor new classes of attacks, high false alarm and so on. In this research work, we have tried to overcome the above stated problems. We have introduced an Intrusion Detection System (IDS) using multiple Machine Learning (ML) Classifier techniques on Message Queuing Telemetry Transport - Internet of Things - Intrusion Detection System dataset (MQTT-IoT-IDS2020) for identifying the multi-class intrusion attacks in the Internet of Things (IoT). In this research work, a result analysis comparison has also been performed to evaluate the performance of all implemented ML classifier techniques. The experiments result obtained higher rates of accuracy. The overall accuracy of experimented Intrusion Detection System has 97.76%, 97.80%, 97.58%, 99.98%, 99.98%, and 97.58% using k-Nearest Neighbour (k-NN), Support Vector Machine (SVM), Naive Bayes (NB), Random Forest (RF), Decision Tree (DT) and Stochastic Gradient Descent (SGD) classifiers respectively. Precision, Recall, and F1-Score of all experimented ML classifiers are mentioned in the result and analysis section.