Secrecy performance intelligent prediction for CRNs: An Self-CondenseNet approach

Abstract

This paper studies the physical layer security (PLS) of cognitive radio networks (CRNs) with Fisher–Snedecor $F$ distribution. To resolve the security issues within CRNs, we derived exact expressions of the security outage probability (SOP) and the probability of strict positive secrecy capacity (SPSC) for the first time, where the SOP and SPSC are uniformly given by Meijer’s G-function. The correctness of the theoretical derivations is proved by Monte Carlo simulations. The results indicate that reducing $m$ and increasing the ratio of $(ms,{\mu }_E)$ will reduce SOP and increase SPSC. Moreover, we proposed the Self-CondenseNet model to predict the security performance of the system. By comparing with three deep learning algorithms of Transformer, MLP-Mixer and CondenseNet, the results show that the proposed Self-CondenseNet has the best prediction performance. Compared with the CondenseNet, the proposed Self-CondenseNet has a 78.26% higher accuracy and a 12.86% lower time complexity. Compared with the MLP-Mixer, the proposed Self-CondenseNet has a 85.29% higher accuracy. The comparison results show that the proposed algorithm has high prediction accuracy and low time complexity, and can be widely used in complex and changeable scenarios such as 5G, Internet of Vehicles (IoV), and mobile vehicle networking .etc.