A Digital Power Grid Information Security Protection Method Based on Federated Learning and Deep Learning

A digital power grid information security protection method based on federated learning and deep learning is proposed. Firstly, an overall architecture for digital power grid information security protection was constructed to achieve rapid and reliable processing of massive power grid information. Then, the BiLSTM-DNN model is formed by integrating bidirectional long short-term memory network (BiLSTM) and deep neural network (DNN), and the improved BiLSTM-DNN model is obtained by utilizing multi head attention mechanism. Finally, the improved BiLSTM-DNN model will be used as the local and global models of the federated learning architecture, and the attack detection results of the local models will be aggregated to achieve security protection of digital power grid information. Based on the NSL-KDD dataset, the proposed method was experimentally demonstrated, and the results showed that its attack detection accuracy and protection success rate were 98.15 % and 97.92 %.