Real-Time Intrusion Detection For IIOT: Advancing Edge Computing Security with Machine Learning-Based Solutions

The emergence of IIoT in vital industrial systems has heightened the demand for strong security frameworks as cyber threats become more sophisticated. However, the traditional security measures are usually inadequate, especially in real-time operational scenarios, making the industrial systems susceptible to disruptive attacks. This is the problem that this paper tries to solve by proposing a new Intrusion Detection System (IDS) for IIoT that uses Machine Learning (ML) and Deep Learning (DL) approaches, supplemented by Neighborhood Component Analysis (NCA) for improved feature selection. Our holistic method is based on pre-processing the dataset, which includes one-hot encoding and min-max normalization, applying NCA for the most critical features selection, and using a two-level classification strategy, where ensemble classifiers, standard classifiers, and deep neural networks are used. IIoT based IDS shows enhanced performance with the use of NCA in feature selection. NCA-optimized models deliver near perfect values of accuracy, precision, recall, F1-score, and AUC, which are close to 1, outperforming Without-NCA. Moreover, the NCA usage decreases both training and testing times significantly, making the system much more effective for real-time applications.