An Empirical Evaluation of Machine Learning Algorithms for Intrusion Detection in IIoT Networks

Industry’s reliance on IoT has skyrocketed as the technology has developed and has transformed into the Industrial IoT (IIoT). The IIoT has brought significant benefits to the industry, including increased safety, efficiency, etc. These advancements are accompanied by security challenges such as large exposed attack areas in IIoT devices, therefore it has become increasingly important to use mechanisms to detect any attack or breach on the IIoT networks. An increasing corpus of research acknowledges the value of machine learning methods for anomaly detection. This paper highlights the importance of machine learning in anomaly detection in IIoT networks, provides a literature review of the implementations of ML algorithms in IIoT networks in the last decade, and finally a comparative study of ML algorithms for anomaly detection in IIoT network data. Support Vector Machine (SVM), Linear Regression (LR), K-nearest neighbor (KNN), and Random Forest (RF) are only a few of the ML techniques utilized in comparative analysis. Results of the comparative analysis show that RF algorithm works best for anomaly detection and the results of RF and KNN are comparable in multi-class classification.