Isolation Forest and Spectral Clustering Based on Cooperative Spectrum Sensing Against SSDF Attack in CWSNs

Cognitive radio (CR) is an effective solution to address the scarcity of wireless communication spectrum, and cooperative spectrum sensing (CSS) can overcome the detrimental effects of channel conditions in single-node detection. However, CSS is vulnerable to spectrum sensing data falsification (SSDF) attack launched by malicious sensor nodes (MSNs). Therefore, identifying MSNs in cognitive wireless sensor networks (CWSNs) is crucial for improving the detection performance of CSS. This paper proposes an isolation forest and spectral clustering (IFSC)-based fusion algorithm for MSN detection, which combines the advantages of anomaly detection and clustering. IFSC estimates the proportion of MSNs using isolation forest, determines algorithm parameters, and develops dynamic defense strategies to adapt to varying attack intensities. Furthermore, spectral clustering assisted by anomaly scores can distinguish between MSNs and normal sensor nodes (NSNs) without labeled data. Simulation results demonstrate the effectiveness of IFSC in defending against attacks under different intensities and frequent node changes. Compared to existing algorithms, IFSC requires only a small size of samples to achieve high sensing accuracy, and its response time is more advantageous in large-scale networks.