Machine Learning Techniques Based on Primary User Emulation Detection in Mobile Cognitive Radio Networks.

Area of Science:

• Wireless Communication
• Cognitive Radio Networks
• Machine Learning Applications

Background:

• Spectrum scarcity in 4G/5G networks necessitates alternative mobile communication like Mobile Cognitive Radio Networks (MCRNs).
• MCRNs utilize spectrum holes from Primary Users (PUs), requiring effective spectrum sensing to avoid interference.
• Primary User Emulation (PUE) attacks pose a threat by mimicking PU signals to gain unauthorized spectrum access.

Purpose of the Study:

• To investigate the efficacy of machine learning techniques in detecting Primary User Emulation (PUE) attacks.
• To compare the performance of Support Vector Machine (SVM), Random Forest, and K-Nearest Neighbors (KNN) algorithms in identifying PUE.
• To evaluate detection accuracy under varying signal-to-noise ratio (SNR) conditions using a Software-Defined Radio (SDR) testbed.

Main Methods:

• Implementation of machine learning classification algorithms: SVM, Random Forest, and KNN.
• Conducting simulation and emulation experiments on a Software-Defined Radio (SDR) testbed.
• Analyzing the probability of PUE detection across different algorithms and SNR levels.

Main Results:

• The Support Vector Machine (SVM) technique demonstrated superior performance in PUE detection.
• SVM achieved an 8% higher probability of detection compared to the energy detector at low SNR.
• SVM outperformed KNN and Random Forest by 5% in PUE detection during experiments.

Conclusions:

• Machine learning, particularly SVM, offers a robust solution for detecting PUE attacks in MCRNs.
• SVM provides enhanced PUE detection capabilities, especially in challenging low SNR environments.
• The study validates the practical application of ML algorithms on SDR testbeds for cognitive radio security.