A Deep Learning-Based Method for Intrusion Detection in Smart Grid: A Case Study of Distributed Denial of Service Detection

Smart grids facilitate automatic control and management of the power system through cyber systems, which include digital devices and the Internet of Things. However, the integration of cyber systems into traditional grids introduces the risk of cyber-attacks, which adversely affect both physical and cyber systems. Distributed denial of service (DDoS) attacks are among the most common cyber securities problems that disturb power flow and, therefore, increase economic losses. In this regard, this paper presents a deep learning-based method to deal with DDoS attacks in the smart grid. First, the important features are selected through correlation analysis. Then, a convolutional neural network (CNN) model is utilized to obtain dependencies between features, and a bidirectional long short-term memory (BiLSTM) model is implemented to extract long-term dependencies between extracted features. Finally, the extracted features of CNN and BiLSTM are used for DDoS detection. According to comparative analysis, the accuracy and F1 score of the presented method increased compared to the traditional and advanced machine learning methods such as logistic regression, feed-forward neural network, multi-layer perceptron, CNN, and BiLSTM models.