A Novel Semi-Supervised Framework for Cybersecurity Threat Detection in Wireless Sensor Networks

In this work, we present a semi-supervised learning framework designed to detect four major categories of attacks in Wireless Sensor Networks (WSNs): Denial of Service (DoS), Probe, Remote-to-Local (R2L), and User-to-Root (U2R). The framework combines the strengths of both supervised and unsupervised learning, using Support Vector Machines (SVM) for classification and Density-Based Spatial Clustering of Applications with Noise (DBSCAN) for clustering. We evaluated the proposed model on the NSL-KDD dataset, where it demonstrated strong performance in terms of accuracy and F1-score. Our analysis also explored how variations in DBSCAN parameters influence detection outcomes, emphasizing that careful parameter tuning is essential for achieving optimal performance. One of the key benefits of this semi-supervised approach is its ability to effectively handle large volumes of unlabeled data—a limitation often encountered when relying solely on supervised or unsupervised methods. By making efficient use of the available labeled data and incorporating clustering techniques, the model delivers improved accuracy and robustness in intrusion detection for WSNs. In summary, this research contributes to the advancement of intrusion detection systems in WSNs by proposing a practical and efficient semi-supervised framework. The findings highlight not only enhanced detection across multiple attack types but also provide valuable guidance on parameter optimization and effective dataset utilization.