Outage Probability Minimization in Secure NOMA Cognitive Radio Systems with UAV Relay: A Machine Learning Approach

• Published in: IEEE transactions on cognitive communications and networking Vol. 9; no. 2; p. 1
• Main Authors: Vo, Van Nhan, Nguyen, Le-Mai-Duyen, Tran, Hung, Dang, Viet-Hung, Niyato, Dusit, Cuong, Dang Ngoc, Luong, Nguyen Cong, So-In, Chakchai
• Format: Journal Article
• Language: English
• Published: Piscataway IEEE 01.04.2023; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Cognitive radio; constrained continuous genetic algorithm; Constraints; Eavesdropping; Genetic algorithms; Machine learning; MIMO communication; non-orthogonal multiple access; Nonorthogonal multiple access; Optimization; Outages; Probability; Relay; unmanned aerial vehicle; Unmanned aerial vehicles
• ISSN: 2332-7731; 2332-7731
• DOI: 10.1109/TCCN.2022.3226184

This paper considers a multiple-input multiple-output (MIMO) non-orthogonal multiple access (NOMA) cognitive radio (CR) system with an unmanned aerial vehicle relay (UR). In this system, a secondary transmitter (ST) uses licensed spectrum from the primary network to transmit signals to its secondary receivers (SRs) based on NOMA. The UR is used as a relay to forward the signals from the ST to the SRs. As a result, the system can achieve significant improvements in spectral efficiency and network capacity. However, such a MIMO NOMA CR system faces issues of interference and security, i.e., eavesdropping attacks, due to the shared spectrum use and the UR. Therefore, we aim to minimize the outage probability of the secondary network, subject to constraints on the outage probability of the primary network and the intercept probabilities of eavesdroppers. Then, we attempt to optimize the transmit power of the UR, the coordinates of the UR, and the power allocation factors for NOMA. We further derive closed-form expressions for the outage probabilities of the secondary and primary networks and the intercept probabilities at the eavesdroppers. We propose using a machine learning algorithm based on a constrained continuous genetic algorithm to solve the optimization problem.