Res-CLN: security intrusion detection system for optical communication in hybrid networks with CNN-LSTM based on fused residual structure

With the development of technologies such as all-optical networks, fiber-to-the-home, and automatic switching networks, optical fiber communication is experiencing a new peak. However, eavesdropping techniques targeting optical fiber communication signals are also becoming increasingly sophisticated. Additionally, in the high-speed data transmission and processing environment of all-optical networks, optical communication faces potential hazards where even brief or commonly patterned attacks could lead to significant data corruption or leakage. Therefore, the security and privacy protection of optical fiber communication cannot be overlooked. Designing an intrusion detection system capable of accurately and promptly capturing malicious traffic and taking preventive measures is crucial for maintaining the security of optical communication networks. Currently, many researchers are using machine learning and deep learning for intrusion detection. However, these methods have some drawbacks, such as manual feature selection, inability to automatically extract spatial and temporal features from traffic data, and relatively shallow network layers. To address these limitations of intrusion detection methods, we propose a model called Res-GLN. It combines Convolutional Neural Networks (CNNs) and Long Short-Term Memory (LSTM) networks into a hybrid network to detect abnormal traffic in networks. It is worth noting that we designed two residual network blocks to deepen the hierarchical structure of the combined CNN and LSTM network, enabling the model to learn better feature representations without compromising performance. We conducted extensive experiments on three public network traffic datasets, and compared to other baselines, Res-GLN exhibits higher accuracy and lower false alarm rates.