Security detection method of illegal intrusion in optical transmission sensor networks based on a dual-path refinement attention mechanism

Abstract

This paper proposes a security detection method based on a dual-path refined attention mechanism to address the complex and variable illegal intrusion behaviours and low detection efficiency in optical transmission sensor networks. By constructing a multi-scale convolutional attention module, efficient fusion of local and global features can be achieved, combining ResNet50 structure optimisation with XGBoost classifier to improve the model ability to discriminate intrusion features and detection speed. The experimental results demonstrate that this method outperforms existing methods in terms of detection rate, real-time performance, and anti-interference capability, particularly in maintaining low network overhead under high attack densities. This study provides a reliable intrusion detection solution for optical transmission sensor networks, which is of great value in enhancing security protection capabilities in complex network environments.