Analysis and Prediction of Mobile Industrial Internet of Things (IIoT) Communications Based on FL-GLP-Net

The number of mobile users and applications is rising quickly due to the deployment of fifth-generation (5G) communication technology. The prevalence of smart devices and Internet of Things (IoT) services has made this technology essential to manage the resulting volume of data. However, the increase in mobile communications raises security concerns considering the open and dynamic nature of the Industrial IoT (IIoT) environment. An important research problem is how to exploit the characteristics of wireless channels for safe and reliable information transmission. Therefore, a mobile communication performance analysis and prediction algorithm based on FL-GLP-Net is proposed. First, a mobile security communication system model based on N-Nakagami channels is presented. Then, the nonzero secrecy capacity probability (NSCP) is studied and an exact expression is derived. XGBoost is used to choose the best features based on model performance. A real-time mobile security performance prediction model based on FL-GLP-Net is designed for real-time NSCP prediction. Federated learning (FL), graph attention network (GAT), long short-term memory (LSTM), and pyramid visual converter (PVT) modules are integrated to obtain a model that can deal with the diverse signal features in mobile communication systems. Results are presented to show that the proposed method outperforms other NSCP prediction algorithms. In particular, the mean-squared error (MSE) of FL-GLP-Net is 78% better than that of FL-ShuffleNetV2.