CNN-LSTM-based Satellite Identity Authentication Design for Mega-Constellation Systems

As mega-constellation systems become increasingly prevalent in both commercial aerospace and global communication infrastructure, robust satellite identity authentication mechanisms have emerged as a critical requirement. In this paper, we design an identity verification scheme that synergistically combines Time-Difference-of-Arrival (TDoA) measurements with deep learning techniques. In particular, our design firstly leverages ground station (GS) receptions of satellite signals to construct TDoA matrices, then the Convolutional Neural Network (CNN) and Long Short-Term Memory (LSTM) networks are utilized to extract temporal and spatial correlations within TDoA measurements. Finally, based on the trained CNN-LSTM model, the high-precision satellite identification classification is achieved. Comprehensive simulations demonstrate that our design outperforms the benchmark scheme in terms of accuracy and response speed, while maintaining computational efficiency. In addition, the simulations also indicate that with optimal configuration of GSs and observation periods, our design delivers exceptional recognition accuracy, robust performance, and strong generalization capabilities, thus making it particularly suitable for secure access control in mega-constellation satellite systems.