Deep Learning-Based Secure Communication for IRS-Aided Terahertz Systems

View at publisher Summarise with VIRAT
Authors: Mingyu Ma, Jingping Qiao, Chuanting Zhang, Haixia Zhang
Cited by: 0
References: 16

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

This paper focuses on physical layer security in an intelligent reflecting surface (IRS)-aided terahertz multiple-input multiple-output (MIMO) systems, where a multi-antenna eavesdropper may eavesdrop either the base station (BS)-to-IRS or IRS-to-user links. With imperfect eavesdropper channel state information (ECSI), we formulate a joint optimization problem for BS active beamforming and phase-shift matrix design to maximize the worst-case secrecy rate. To address the high computational complexity of conventional convex optimization-based methods, we present an efficient deep learning (DL)-based approach that jointly optimizes the BS active beamforming and IRS phase-shift matrix for both potential eavesdropping channels. Numerical simulations verify that the proposed DL method maintains comparable security performance while reducing computational complexity up to an order of magnitude compared to traditional alternating optimization algorithms, making it particularly suitable for real-time implementation in terahertz-band communication systems.

Fields of Study & Subjects

Wireless Communication Security Techniques Advanced Wireless Communication Technologies Millimeter-Wave Propagation and Modeling

Publication details

Journal / source
IEEE Wireless Communications Letters
Volume
15
Pages
2453–2457
Publication date
1 January 2026
Language
EN