Stacking Ensemble-Based Framework for Intrusion Detection in Internet-of-Energy (IoE) Systems

Internet-of-Energy (IoE) development has been the result of the rapid growth of ICT coupled with the increase in demand for power systems. Such systems integrate IoT devices with ICTs which enable real-time energy management. Crucial components of IoE systems are Distributed energy resources (DERs) and battery energy storage systems (BESS). However, this makes IoE systems vulnerable to a variety of cyber attacks that can have dire consequences such as DDoS, SQL injection, and false data injection. Machine learning (ML) approaches have shown great promise in detecting attacks in networked environments, but may sometimes have bias/variance. To that end, this work proposes an ensemble-based framework that considers three different stacking ensemble approaches built using tree-dependent models. This choice aims to improve the overall performance and generalize well while simultaneously having reduced model bias/variance. Experimental results illustrate that the proposed stacking ensemble framework achieves better performance in terms of accuracy, precision, recall, and F1-score while having comparable training and prediction speeds. This demonstrates the framework's advantage from both a detection capability and computational/inference efficiency perspective.