Modelling for Cognitive Radio Networks using Deep Learning and Optimization Techniques
M. Saraswathi, E. Logashanmugam · 2024
Malicious user detection for spectrum sensing (SS) in Cognitive Radio Networks (CRNs) is a crucial safety component to ensure these systems operate proficiently and reliably. Spectrum sensing allows CRNs to locate and utilize available spectrum bands, but it is also vulnerable to interference and malicious behavior. Classifying harmful users is essential for maintaining network integrity. Deep learning (DL)-based malicious user association leverages advanced neural network frameworks to detect and highpoint potential vulnerabilities within a network. By analyzing large volumes of data, DL techniques can identify patterns and anomalies associated with spiteful user behavior, such as system intrusions, fraud, or abnormal actions. This study offering a Deep Belief Network (DBN) with Sand Cat Swarm Optimization (SCSO)-based spectrum method that combines both DBN and SCSO to optimize spectrum utilization. DBNs are proficient at detecting signal appearances, enabling them to excerpt essential features from IQ-based spectrum data. Modifier networks, meanwhile, account for long-term dependencies, allowing these features to be further enhanced and refined. This method captures both local and global patterns in the spectrum data, resulting in reduced sensing errors and a higher probability of detection. By incorporating DBN's deep learning abilities with SCSO's nature-inspired measures, a synergistic framework is created that allows CRNs to autonomously detect and assign incidents with significant precision. The proposed approach significantly enhances spectrum capability and the robustness of CRNs by leveraging the power of DBN, leading to more effective resource utilization and summary interference. The performance of the suggested model is evaluated in terms of prediction error and efficiency. The GSM-900 spectrum dataset demonstrates that DBN-SCSO outperforms existing state-of-the-art prediction procedures in terms of prediction spectrum efficiency of 94.78%, throughput of 1577.73kbps and accuracy of 96.78%.