Achieving Cyber Resilience in 5G-Enabled Microgrids Using Supervised Learning

Achieving Cyber Resilience in 5G-Enabled Microgrids Using Supervised Learning

연구 분야: Infrastructure

학회: World Congress in Computer Science, Computer Engineering & Applied Computing

Marine Corps Air Station (MCAS) Miramar is the most energy forward defense installation in the U.S. The base has implemented a comprehensive networked microgrid that leverages 5G for energy resource communications and energy asset monitoring. However, the 5G implementation at Miramar is Non-Stand Alone (NSA), meaning it continues to anchor to the 4G core network, while using 5G frequencies. The 5G NSA set up has limitations compared to the 5G standalone architecture, particularly in its ability to detect cyber anomalies in energy traffic. As a first step in minimizing this vulnerability, this research aims to develop a network traffic anomaly detection model using supervised machine learning (ML). We use a Long Short Term Memory (LSTM) and Support Vector Machine (SVM) to identify 5G specific anomalies that may occur within the energy communications infrastructure at MCAS Miramar. We train these models on five different feature representations of simulated 5G network traffic dataset. Our results compare and contrast the SVM and LSTM and show that the SVM model generated highly satisfactory results in terms of accuracy and precision of nearly 99%. The utilization of these ML models enhance MCAS Miramar’s energy security and resilience.