Main article

Minh Anh Le
Department of Telecommunications Engineering, University of Transport and Communications, Hanoi, Vietnam
Quang Bao Tran
Faculty of Electronics and Telecommunications, Industrial University of Ho Chi Minh City, Ho Chi Minh City, Vietnam
Thao Linh Pham*
Faculty of Information Technology, Hanoi University of Industry, Hanoi, Vietnam
thao.linh.pham@haui.edu.vn

DOI: https://doi.org/10.63646/

Abstract

Secure cognitive radio networks require rapid decisions about relay use and secondary-user power while protecting the quality of service of the primary network and limiting the probability that an eavesdropper can decode secondary data. Existing analytical studies of multihop underlay relaying have clarified the outage and interception behavior of TAS/SC-assisted cooperative protocols under generalized fading, but their decision logic is usually fixed before deployment and depends on simplified assumptions about channel state availability, node mobility, and eavesdropper position. This article develops an AI-enabled relay and power selection framework for secure cognitive radio networks. The framework formulates relay activation, antenna-combining mode selection, and secondary power adjustment as a constrained learning problem in which the agent observes channel-quality, interference-margin, hop-distance, queue, and security-risk states and then selects safe actions under a primary-network outage guard. A simulation study is constructed around a multihop underlay scenario with MIMO secondary nodes, incremental decode-and-forward cooperation, and a passive multi-antenna eavesdropper. The proposed safe reinforcement learning policy is compared with fixed-rule, greedy-reliability, and security-prioritized baselines. Across the benchmark settings, the learning policy reduces end-to-end outage by 38.6% against the greedy baseline and 63.1% against the fixed-rule baseline, while holding interception probability close to the security-prioritized policy. The article contributes a practical decision architecture, an interpretable reward design, and a data-analysis template for evaluating reliability-interception trade-offs without relying on heavy closed-form derivations.

Article details

How to Cite

Le, M. A., Tran, Q. B., & Pham, T. L. (2025). AI-Enabled Relay and Power Selection for Secure Cognitive Radio Networks: Learning-Based Optimization of Reliability and Interception Risk. Journal of AI Analytics and Applications, 3(2), 50-69. https://doi.org/10.63646/