Machine Learning Algorithms to Defend Against Routing Attacks on the Internet of Things: A Systematic Literature Review

Abstract

The Internet of Things (IoT) has become increasingly popular, opening vast application possibilities in different fields including smart cities, healthcare, manufacturing, agriculture, etc. IoT comprises resource-constrained devices deployed in Low Power and Lossy Networks (LLNs). To satisfy the routing requirements of these networks, the Internet Engineering Task Force (IETF) created a standardised Routing Protocol for low-power and Lossy Networks (RPL). However, this routing protocol is vulnerable to routing attacks, prompting researchers to propose several techniques to defend the network against such attacks. Machine learning approaches demonstrate effective ways to detect such attacks in large quantities. Therefore, this paper systematically synthesised 17 publications to compare the performance of traditional and advanced machine learning algorithms to identify the best algorithm for detecting RPL-based IoT routing attacks. The findings of this paper show that machine learning algorithms are capable of effective detection of many routing attacks with high accuracy and a low False Positive Rate. Furthermore, the results demonstrate that on average, advanced machine learning algorithms can achieve an accuracy of 96.03% compared to traditional machine learning algorithms which achieved 91.67%. Traditional machine learning algorithms demonstrated the best performance on average False Positive Rate by achieving 2.75% compared to their counterparts which gained 4.79%. However, Random Forest showed the best performance and outperformed all the algorithms in the selected publications by achieving over 99% accuracy, precision and recall.

Keywords

• rpl
• iot
• lnns
• machine learning
• routing attacks