An Anomaly Detection Algorithm Based on Ensemble Learning for 5G Environment.

Area of Science:

• Computer Science
• Network Security
• Artificial Intelligence

Background:

• The proliferation of digital services (VR, IoT, cloud) drives demand for 5G bearer networks with high capacity and advanced features.
• Software-defined networking (SDN) offers flexibility for 5G network slicing but faces significant network intrusion challenges.

Purpose of the Study:

• To develop a robust abnormal traffic detection method for software-defined networking environments.
• To address the limitations of traditional ensemble learning in capturing long-term data dependencies.

Main Methods:

• A novel stacking ensemble method is proposed, integrating a self-attention mechanism with a convolutional network.
• This approach learns long-term associations between traffic samples, generating embeddings for downstream tasks.
• A fusion module combines sample embeddings and base learner predictions for final outlier detection.

Main Results:

• The proposed method achieved high performance on abnormal traffic datasets in SDN environments.
• Precision, recall, and F1-score reached 0.9972, 0.9996, and 0.9984, respectively.
• These results surpass those of existing comparison algorithms.

Conclusions:

• The developed stacking ensemble method with self-attention effectively detects abnormal traffic in SDN.
• This technique enhances network security by accurately identifying intrusions.
• The findings offer a significant improvement for securing next-generation networks like 5G.