Efficient Cyber Attack Detection in Wireless Sensor Networks Using ANOVA F-Test and XGBoost

Wireless Sensor Networks (WSNs) are indispensable in many different applications. Hence, reliability and security of these systems depend on effective detection systems. In this work, we use XGBoost for classification and the ANOVA F-Test for feature selection to provide a model for cyber threat detection in WSNs. We picked the best features using ANOVA F-Test. Particularly excelling in identifying TDMA and Blackhole attack types, our XGBoost model obtained a 90% accuracy, surpassing Logistic Regression, SVM, Gradient Boosting, and Catboost in recall and F1-score across most attack classes.