Malicious user classification in cognitive 5G networks using novel improved bidirectional encoder representations...

Area of Science:

• Cybersecurity
• Wireless Communication
• Machine Learning

Background:

• Identifying malicious users is crucial for securing cognitive 5G networks against dynamic spectrum access attacks.
• Challenges include network complexity, limited labeled data, and evolving attack vectors, necessitating adaptable detection methods.

Purpose of the Study:

• To develop and evaluate a novel intelligent machine learning methodology for Malicious User Classification in Cognitive 5G Networks (MUC-C5GN).
• To enhance real-time detection capabilities, reduce false alarms, and improve generalization in dynamic 5G environments.

Main Methods:

• Utilized the 5G Network Intrusion Detection Dataset (5G-NIDD) for data collection.
• Employed Self-Attention Recurrent Neural Network-Autoencoder (RNN-AE) for feature extraction.
• Implemented an Improved Bidirectional Encoder Representations from Transformers (IBERT) model optimized by the Revolution Optimization Algorithm (ROA) for classification.

Main Results:

• The proposed IBERT-ROA model achieved superior performance with 99.74% accuracy, 98.48% sensitivity, and 98.91% precision.
• Demonstrated significant improvements over existing methods, including up to 5.99% in sensitivity and 2.74% in accuracy.
• Effectively classified seven types of attacks and benign traffic.

Conclusions:

• The IBERT-ROA methodology offers an effective, scalable, and suitable solution for real-time malicious user detection in cognitive 5G networks.
• The approach provides dependable effectiveness and confidence in cognitive radio-enabled 5G communication frameworks.