Random forest-based IDS for IIoT edge computing security using ensemble learning for dimensionality reduction
Mouaad Mohy-eddine, Said Benkirane, Azidine Guezzaz, Mourade Azrour
Abstract
The industrial internet of things (IIoT) is the IoT application in the industrial sphere. It aims to get smart devices involved in industrial sectors to enhance their operations. In addition to IoT vulnerabilities, IIoT introduces more severe security flaws. Therefore, intrusion detection systems (IDS) are being developed to avoid catastrophic intrusions. This research proposes a machine learning (ML)-based IDS for IIoT edge computing security. We employed Pearson's correlation coefficient (PCC) and isolation forest (IF) to reduce the computational cost and training. We implemented the IF to eliminate outliers and the PCC for the feature selection. Feature engineering improves ML models' accuracy (ACC) and detection rate (DR). The random forest (RF) classifier was applied to enhance the IDS performances. Our approach showed significant results on the Bot-IoT dataset with 100% DR and 99.99% ACC. The obtained results show that our approach has many advantages compared to other models.