Enhanced Reinforcement Learning with Optimization for the Threat Detection in VLC Networks-based 5G

Abstract: Visible Light Communication (VLC) offers significant bandwidth, energy efficiency, and resistance to electromagnetic interference, making it a promising addition to 5G technology. However, due to its broadcast nature and dependence on Line-of-Sight (LoS), it is vulnerable to threats such as eavesdropping, spoofing, and jamming. The unique characteristics of optical signals, background noise, and the need for real-time, low-latency communication make threat detection in VLC-based 5G networks challenging. Therefore, this research identifies threats in VLC-based 5G networks using a novel intelligent, machine-learning-based optimization approach. Threat data are collected from the benchmark VLCdata dataset and preprocessed using the Discrete Wavelet Transform (DWT) denoising technique. Feature extraction is then performed using the Hough Transform (HT). Finally, threats are detected using the proposed Enhanced Reinforcement Learning (ERL) model, in which the Builder Optimization Algorithm (BOA) is employed to optimize the reinforcement learning parameters. The optimization process aims to maximize detection accuracy. Experimental results demonstrate that the proposed ERL-BOA model outperforms existing approaches across multiple performance measures. Specifically, it achieves 8.76% higher accuracy and 4.17% higher F1-score than existing methods, demonstrating its effectiveness for intelligent threat detection in VLC-based 5G networks.

Keywords: Visible Light Communication Networks; 5G; Discrete Wavelet Transform (DWT) Denoising; Hough Transform; Enhanced Reinforcement Learning; Builder Optimization Algorithm.