Intelligent Warning Technology for Password Threats in Energy Information Systems Based on the BiMAGRU-BiSALSTM Model

With the acceleration of the informationisation process in the energy industry, password threats in information systems have become an important hidden threat affecting their operational security. An intelligent warning technology based on the Bi-MAGRU Bi-SALSTM model is proposed to address the challenges of password threat detection and classification. The innovative use of hash functions for data transformation in the data pre-processing stage and the combination of mutual information scores to extract key features have improved the effectiveness and expressiveness of data input. In terms of model design, a classifier model specifically designed for information systems in the energy industry has been constructed, which combines a Bidirectional Gated Recurrent Unit (Bi-GRU) encoding layer with a multi-head attention mechanism and a Bidirectional Long Short-Term Memory (Bi-LSTM) decoding layer to achieve deep mining of temporal correlations and global features in encrypted traffic. Extensive experiments have demonstrated that this method substantially surpasses current techniques in classification accuracy, detection efficiency, and robustness. On the TOR-NOTOR dataset, the F1 score increased by 3.70%, and the accuracy by 116.72% relative to the latest advanced methods; on the Encrypted Malicious Traffic Detection dataset, the accuracy rose by 7.11%.