Secure Cognitive Radio Communication via Intelligent Reflecting Surface

TL;DR

This paper investigates how intelligent reflecting surfaces can improve the secrecy rate in cognitive radio channels with eavesdroppers, proposing new optimization solutions for various levels of eavesdropper CSI and demonstrating significant performance gains.

Contribution

It introduces novel numerical algorithms for optimizing secrecy rates in IRS-assisted cognitive radio channels considering different eavesdropper CSI scenarios, addressing limitations of existing solutions.

Findings

Proposed algorithms significantly improve secrecy performance.

Positive secrecy rate achievable even without eavesdropper CSI.

AO algorithms guarantee monotonic convergence.

Abstract

In this paper, an intelligent reflecting surface (IRS) assisted spectrum sharing underlay cognitive radio (CR) wiretap channel (WTC) is studied, and we aim at enhancing the secrecy rate of secondary user in this channel subject to total power constraint at secondary transmitter (ST), interference power constraint (IPC) at primary receiver (PR) as well as unit modulus constraint at IRS. Due to extra IPC and eavesdropper (Eve) are considered, all the existing solutions for enhancing secrecy rate of IRS-assisted non-CR WTC as well as enhancing transmission rate in IRS-assisted CR channel without eavesdropper fail in this work. Therefore, we propose new numerical solutions to optimize the secrecy rate of this channel under full primary, secondary users' channel state information (CSI) and three different cases of Eve's CSI: full CSI, imperfect CSI with bounded estimation error, and no CSI. Simulation results show that our proposed solutions for the IRS-assisted design greatly enhance the secrecy performance compared with the existing numerical solutions with and without IRS under full and imperfect Eve's CSI. And positive secrecy rate can be achieved by our proposed AN aided approach given most channel realizations under no Eve's CSI case so that secure communication also can be guaranteed. All of the proposed AO algorithms are guaranteed to monotonic convergence.