A contrived dataset of substation automation for cybersecurity research in the smart grid networks based on IEC61850

Abstract

Relevant datasets that depict and/or emulate the smart grid networks (SGN) are key to developing cybersecurity models that can effectively provide security for mission-critical infrastructure. The difficulty in obtaining relevant SGN real-life datasets presents a considerable challenge for researchers in the field and, the existing datasets lack representation of the IEC61850 protocol for modelling substation automation processes for cybersecurity solutions. This paper presents a dataset simulated from a fully virtual testbed, intended to provide researchers with the necessary datasets for research and experiments that require massive amounts of data close to the real-world scenario. Experimentally, the dataset was used to develop an intrusion detection model based on gated recurrent unit (GRU), deep belief network (DBN), long-short term memory (LSTM), and evaluated using accuracy, precision, recall, F1-score, detection rate, false alarm rate (FAR), missed alarm rate (MAR), mean squared error (MSE), mean absolute error (MAE), and loss. Results show that the standalone deep learning (DL) algorithms outperformed the hybridized ones and are more suitable for developing models for securing substation automation systems that support generic object-oriented substation events (GOOSE), and manufacturing message specification (MMS) and run on IEC61850, distributed network protocol version 3 (DNP3), and Modbus-transmission control protocol (ModbusTCP).