Secure Performance Analysis of Aerial RIS-NOMA-Aided Systems: Deep Neural Network Approach

Abstract

The next generation of wireless systems has benefits in terms of spectrum and energy inefficiencies by exploiting two promising techniques including Non-Orthogonal Multiple Access (NOMA) and Reconfigurable Intelligent Surfaces (RIS). The scenario of two legitimate users existing together with an eavesdropper is worth examining in terms of secure matter while enabling machine learning tools at the base station for expected improvement. The base station deals with a highly complicated algorithm to adjust parameters against the attack of eavesdroppers and to better improve the secure performance of mobile users. This paper suggests a better solution to allow the base station to predict performance at destinations to adjust necessary parameters such as power allocation coefficients properly. To this end, we propose a deep neural network (DNN)-based approach which also leverages the benefits of aerial RIS to achieve predictable performance and significant secure performance improvement could be enhanced. We first derive the formulations for security outage probability (SOP) in closed-form expressions and analyze the strictly positive secrecy capacity (SPSC), which are crucial performance metrics to determine how the systems are against the existence of eavesdroppers. Such eavesdroppers intend to overhear signal transmission dedicated to intended users and incur degraded system performance. The numerical simulations are expected to evaluate how the machine learning tool works with the traditional computation of system performance metrics which is able to be verified by comparing with the Monte-Carlo method. Our numerical simulations demonstrate that the design of a higher number of meta-surface elements at the RIS, as well as a higher signal-to-noise ratio (SNR) levels at the base station, are key parameters to achieving improved security performance for users. For detailed guidelines of the RIS-NOMA aided system, we provide a table of parameters samples resulting in secure performance as expected.