A Novel Prediction Model for Malicious Users Detection and Spectrum Sensing Based on Stacking and Deep Learning.

Area of Science:

• Computer Science
• Electrical Engineering
• Cybersecurity

Background:

• Cooperative spectrum sensing in cognitive radio networks enhances accuracy by sharing sensing data.
• Malicious users in cooperative networks transmit false data, degrading sensing performance and security.
• Existing methods struggle to effectively identify and mitigate interference from malicious users.

Purpose of the Study:

• To propose novel machine learning approaches for detecting malicious users in cooperative spectrum sensing.
• To improve the accuracy and reliability of spectrum sensing decisions in cognitive radio networks.
• To enhance the security and efficiency of cognitive radio networks against malicious attacks.

Main Methods:

• A stacking model for malicious user detection using chaotic compressive sensing for feature extraction and ensemble learning for classification.
• A deep learning technique utilizing scalogram images for primary user spectrum classification.
• Implementation of two distinct machine learning solutions to address security vulnerabilities.

Main Results:

• The stacking model achieved 97% accuracy in detecting malicious users, even with 50% malicious nodes.
• The scalogram-based deep learning technique demonstrated fast spectrum sensing with high detection probability and low false alarm rates.
• Both proposed methods significantly improved the performance and security of cooperative spectrum sensing.

Conclusions:

• The developed machine learning techniques effectively detect malicious users and improve spectrum sensing accuracy in cooperative cognitive radio networks.
• These solutions offer robust security against malicious interference, enhancing the overall reliability of cognitive radio systems.
• The proposed methods represent a significant advancement in securing cooperative spectrum sensing against adversarial attacks.