Intrusion Detection in IoT-based Cyber-physical System ​Using Machine Learning

Abstract

The use of Internet of Things (IoT) devices is a result of the massive number of messages carried through the internet. One of the most serious IoT threats is the botnet attack, which aims to commit legitimate, profitable, and effective cybercrimes. Because of the ramifications of hostile acts in computing, communication, and cyber-physical systems (CPS), researchers are encouraged to develop effective Intrusion Detection Systems (IDSs). Intelligent-powered intrusion detection solutions ideal for deployment in IoT-based cyber-physical settings are created using cutting-edge learning algorithms. In this study, three machine learning (ML) approaches are employed to develop a system that recognizes legal or malicious communication in linked computer networks. The results showed that random forest (RF) fared the best, with a coefficient of determination (R2) of 0.9965 and error scores of 0.0015 for root mean square error (RMSE), 0.0522 for mean absolute error (MAE), and 0.0872 for mean absolute percentage error (MAPE). Despite the models achieved significant results, it was suggested that RF be used when detecting legal or malicious networks in IoT-connected computer networks. This study is crucial for making informed decisions that will lead to long-term growth in industrial and educational networks.

Keywords

• machine learning
• iot
• cps
• intrusion detection
• botnet detection
• network traffic