A simple machine learning technique for sensor network wireless denial-of-service detection

Abstract

Wireless sensor networks (WSNs) are integral to numerous applications but are vulnerable to denial-of-service (DoS) attacks, which can severely compromise their functionality. This research proposes a lightweight machine learning approach to detect DoS attacks in WSNs. Specifically, we investigate the efficacy of decision tree (DT) algorithms with the Gini feature selection method, alongside random forest (RF), extreme gradient boosting (XGBoost), and k-nearest neighbor (KNN) classifiers. Data collected from normal and DoS attack scenarios are preprocessed and used to train these models. Experimental results showcase the effectiveness of the proposed approach, with the DT algorithm exhibiting high accuracy exceeding 90%, surpassing other classifiers in computational efficiency and interpretability. This study contributes to enhancing the security and reliability of WSNs, offering insights into potential future optimizations and algorithmic explorations for robust DoS attack detection.