STEALTH: Space-Time Encryption via Constellation Hiding for MIMO Two-Way Untrusted Relay Systems

Abstract

We propose a novel physical-layer security (PLS) technique for multiple-input multiple-output (MIMO) two-way untrusted relay-assisted (TWUR) internet-of-things (IoT) networks by combining constellation-overlapping-based over-the-air encryption (OTAE) and space-time codes. Like the conventional two-way relay system, the proposed system operates in two phases: multiple access and broadcast. In the first phase, two IoT devices exploit the power-controlled space-time line code (STLC) with truncated channel inversion to transmit data frames such that their constellations completely overlap at the untrusted relay. This makes it impossible for the relay to distinguish the exact signals of the IoT devices, thereby enhancing security. In the second phase, the relay amplifies and forwards the superimposed signal using the space-time block code (STBC). From the superposed signal, each IoT node decodes data transmitted by the other using self-interference cancellation. Simulation results demonstrate that the proposed technique achieves the same security performance as the conventional constellation-overlapping-based OTAE scheme while significantly improving communication performance by effectively leveraging multiple antennas at both the untrusted relay and IoT devices. Specifically, the proposed technique maintains an error floor above 0.1 at the untrusted relay, ensuring secure communication by preventing unauthorized signal decoding. Furthermore, compared to the conventional OTAE scheme, it significantly improves the bit error rate (BER) at IoT devices while maintaining low computational complexity. In contrast to existing MIMO-based PLS schemes, the proposed technique operates without requiring global channel state information (CSI), thereby reducing signaling overhead and making it more practical for IoT networks. Finally, we mathematically analyze the error rate of the proposed technique using the piecewise linear approximation (PLA) and confirm that the derived analytical results closely match the simulation outcomes.