Intrusion Two-Level Detection System Using Machine Learning Techniques for Industrial Internet of Things (IIoT)

The Industrial Internet of Things (IIoT) can be distributed on a large scale with thousands of devices. The massive number of devices can lead to security flows since the IIoT devices have limited resources such as memory and battery. Therefore, there is a need for a fast Intrusion Detection System (IDS) that can identify the behavior of the IIoT network. A proposed model with two detection levels based on machine learning techniques is developed. The first detection level is to detect if the record is normal or an attack. The second detection level in case the record is attack, the type of the attack is determined. The results show that the Decision Tree and Random Forest have the best performance in terms of accuracy. However, the Naive Bayes has the upper hand regarding average processing time with lower accuracy in the first detection level. In the second detection level, the Random Forest outperformed other techniques in terms of accuracy, whereas Naive Bayes still performs the best in average processing time with less accuracy. The proposed model shows high efficiency not only in detecting the attack but also in determining the type of the attack.