Detecting Attacks in the Internet of Things by Utilizing Machine Learning Approaches

Source abstract

Advanced IoT systems have brought great comfort to their users, but at the same time they also increase the risk of being targeted and exploited by malicious cyber attackers. To meet this challenge, empirical analyses show that data pre-processing techniques and classification models of machine learning algorithms used to enhance accuracy can detect attacks with higher performance and classify them as either safe or attacking, compared with traditional technologies that are not effective with booming progress as well as different types of attack and their harshness. Thus, with the aim of accomplishing this goal, We suggest an evaluation of an intrusion detection system (IDS) that utilizes machine learning (ML) techniques with the NSL-KDD dataset. In this research, we have employed three of the most commonly utilized ML classification algorithms within IDS: Decision Tree (DT), Random Forest (RF), and Support Vector Machine (SVM). By thoroughly analyzing the outcomes of the conducted experiments and contrasting them with the findings presented in other research papers. Thus, our experiments using NSL-KDD datasets show that RF and DT are respectively the most accurate. Besides, the results of our experiment also show that SVM suffers the most.