ANALYZING SUPERVISED LEARNING MODELS FOR INTRUSION DETECTION: TOWARDS ROBUST WIRELESS SENSOR NETWORK

The decentralized and resource-constrained nature of Wireless Sensor Networks (WSNs) makes them susceptible to a range of cyberthreats, despite their growing deployment in critical infrastructure. Machine learning-enabled intrusion detection systems (IDS) have become effective instruments for protecting these networks. The models, Random Forest (RF), Decision Tree (DT), Support Vector Machine (SVM), hybrid RF-XGBoost, and semi-supervised techniques, like SVM+DBSCAN, are all evaluated in this papers comparative analysis of various ML-based IDS techniques. In this study, classification performance is measured and compared using the F1-score, accuracy, precision, and recall on the benchmark dataset NSL-KDD. Our findings show that Decision Tree classifiers and hybrid models both attain nearly flawless detection rates, indicating their strong potential for securing Wireless Sensor Networks. This high level of accuracy, combined with low computational overhead, highlights their suitability for real-time intrusion detection in resource-constrained environments. These results reinforce the value of interpretable, lightweight models in practical WSN deployments and mark a significant step forward in achieving robust, scalable network security.