A deep learning based intrusion detection model in 5g environment using lstm network and hyper parameter tuning

In the fifth generation (5G) network environment, high-speed data transmission and low latency are critical. Robust intrusion detection not only helps enhance network security but also aids in maintaining efficient and uninterrupted data flow. Although existing IDS models utilizing hybrid and non-hybrid classification techniques have improved the accuracy of intrusion detection, the complexity of modern network intrusions requires the use of more advanced machine learning methods to analyze complicated network traffic patterns and distinguish subtle anomalies from legitimate behavior, This paper evaluates the effectiveness of Long Short-Term Memory (LSTM) networks in intrusion detection based on the CICIDS2018 dataset, aiming to measure its performance in predicting network intrusion behaviors through metrics such as accuracy, precision, recall, and F1 score. Experimental results indicate that LSTM achieves an accuracy of 98% in identifying intrusion patterns by leveraging the temporal dependence of network traffic features. Furthermore, enabling hyperparameter experiments show that the model maintains stable performance across all evaluation metrics and exhibits excellent robustness under optimized conditions. These findings underscore the LSTM model's potential as a real-time intrusion detection solution in dynamic 5G environments. As network threats become increasingly complex, integrating deep learning technologies like LSTM into the IDS framework can significantly enhance real-time threat mitigation capabilities. This study points out the need for further development of adaptive IDS models to address the evolving network security challenges. Future research should focus on optimizing the computational efficiency and generalization ability of such models to ensure their scalability across diverse network architectures.