A novel multi-dimensional constellation rotation method for physical layer security

This paper introduces a novel multi-dimensional constellation rotation method for physical layer security (PLS) in wireless communication systems. Leveraging an autoencoder (AE)-based deep learning architecture, the proposed scheme jointly learns optimal constellation structures and rotation angles using channel features from both time and frequency domains. The system selects a specific multi-dimensional constellation and applies a rotation based on channel phase, which is unique to each legitimate user. Numerical results show that the proposed system outperforms benchmark systems in bit error rate (BER) while ensuring confidentiality against eavesdroppers. Even with imperfect channel knowledge, the scheme preserves robust security with the eavesdropper’s BER near 0.5. Furthermore, experimental measurements using software-defined radio platforms confirm practical feasibility and demonstrate effectiveness in real-world conditions. These findings highlight the promise of AE-based modulation techniques for strengthening security in next-generation wireless networks.