AI-based intrusion detection for optical wireless communication systems

Abstract

Optical Wireless Communication (OWC) networks provide high-data-rate wireless transmission but are exposed to various cyber threats that can compromise communication reliability and system security. This study proposes a Deep Neural Network (DNN)-based intrusion detection approach for distinguishing legitimate and malicious network traffic. Six communication features, including packet size, transmission delay, signal-to-noise ratio (SNR), received optical power, packet arrival rate, and bit error rate (BER), are utilized as inputs to the classification model. The DNN learns complex feature interactions to identify intrusions with high accuracy and a low false positive rate. Performance evaluation on a dataset comprising normal and attack traffic demonstrates strong classification capability, achieving 99.30 % accuracy, 99.14 % precision, 98.86 % recall, 99.54 % specificity, and a 99.00 % F1-score. The corresponding ROC analysis yields an area under the curve (AUC) of 0.9938, indicating excellent discrimination between benign and malicious traffic. A comparison with existing deep learning techniques shows that the proposed method offers competitive detection performance while preserving computational efficiency and good generalization. The developed framework, therefore, represents a reliable and scalable solution for improving cybersecurity in future OWC environments.