Enhancing Security in DNP3 Communication for Smart Grids: A Segmented Neural Network Approach

The Distributed Network Protocol 3 (DNP3) protocols focus on securing critical infrastructure communication in sectors such as energy and supervisory control and data acquisition (SCADA) systems. The security of DNP3 is paramount, employing features such as encryption, authentication, and secure key management to safeguard against cyber threats. The robust security framework ensures the reliability and integrity of data exchange, fortifying the resilience of industrial control systems against potential cyber-attacks. This study investigates Smart Grid (SG) DNP3 communication security and provides a deep learning-driven approach to detect and prevent cyber-attacks in the SG. Securing communication in SG is a critical challenge, particularly for protocols such as DNP3, which is essential to SCADA systems. This study explores the potential for enhancing intrusion detection in DNP3 communications and the associated industrial control system traffic through the application of state-of-the-art deep learning (DL) algorithms. A Segmented Neural Network (SNN) architecture is employed to analyze the DNP3 dataset, which is captured using CICFlowMeter3 and a DNP3 Parser, integrating Deep Neural Network (DNN), Long Short Term Memory (LSTM), and Random Neural Network (RandNN) models. In CICFlowMeter3, the model achieved an accuracy of 99.86%, whereas, on the DNP3 Parser, it improved to 99.75%, demonstrating outstanding performance. These findings show that the proposed framework is efficient and resilient with complicated and varied data streams. The results show that the proposed SNN-based solution improved the security and resilience of SG operations to detect anomalies in industrial control systems (ICS) in real-time.