Joint Secure Transmission Enhancement of Primary and Secondary Users in RIS Aided Spectrum Sharing Cognitive Radio Networks

Reconfigurable intelligent surface (RIS) has received continuous attentions in recent years due to its unique ability to adjust the amplitude and phase of wireless channels, and it has been widely applied in spectrum sharing (SS)-cognitive radio (CR) networks to enhance the system's secure transmission performance. However, most existing studies only aim to enhance the secure transmission of secondary users (SUs) instead of considering the security issues of both primary users (PUs) and SUs. Motivated by this research defect, this paper aims to jointly enhance the secure transmission performance of PUs and SUs in a RIS aided SS-CR network in the presence of an eavesdropper (Eve). Firstly, we consider an ideal case that the full channel state information (CSI) of all the communication links can be obtained at primary transmitter (PT) and secondary transmitter (ST). Secondly, we consider a more practical case that the primary network and secondary network are non-cooperative in channel estimation and sharing the CSI between each other so that the direct and reflecting communication links between PT and SUs and between ST and PUs are imperfectly known. Furthermore, a strong Eve is considered in this case that can filter out the interference when decoding the confidential information, and the CSI of Eve's related communication links are also imperfectly known at PT and ST. To enhance the secure transmission performance under both cases, we formulate a joint power minimization problem at PT and ST subject to unit modulus constraint at RIS, quality of service constraints at PUs and SUs, interference power constraints at PUs, and eavesdropping rate constraints at Eve under idea case, secrecy outage probability constraints at Eve under practical case. An efficient alternating optimization (AO) in combination with semi-definite relaxation (SDR) algorithm is proposed to solve the problem under ideal case, and an AO in combination with successive convex approximation, SDR as well as penalty convex concave procedure algorithm is proposed to solve the problem under general case. Numerical examples demonstrate the performance advantages of proposed algorithms compared with benchmark scheme.