Terminal Protocol Recognition Method Based on Deep Neural Networks

Network protocols are the foundation for data exchange in power business in the Internet of Things, and accurate and efficient terminal protocol identification is essential. In response to this, this article proposes a protocol identification method that integrates the SENet network model and the bidirectional long short-term memory network model (BiLSTM). The method in this article first uses SENet network to adaptively adjust the network model, focusing on extracting important information from protocol data to achieve accurate extraction of information data. Furthermore, the BiLSTM network is introduced to train the SENet network for processing data features through positive and negative learning, and a complete power terminal protocol sample library is constructed. The simulation experiment uses ISCX2012 data to simulate the complex situation of the power Internet of Things. The result prove that the SENet-BiLSTM protocol recognition method has an precison of over 88% for detecting datasets of various types of protocols, and has excellent ability to identify and analyze terminal protocols.