A Framework for Detecting Low-Power DoS Attacks in Wireless Sensor Networks

Low-Power Denial of Service (LDoS) attacks pose a significant threat to Wireless Sensor Networks (WSNs) because low-rate and intermittent traffic can resemble legitimate network congestion while progressively increasing resource consumption. This study presents an energy-aware framework for evaluating LDoS detection approaches in resource-constrained WSNs by jointly considering detection performance and node energy consumption. A 100-node Zigbee-based star-topology WSN is considered, with Naïve Bayes, Random Forest, and XGBoost evaluated together with threshold/statistical and deep-learning approaches discussed from the literature. The simulation results indicate that lightweight Naïve Bayes achieves a reported detection accuracy of approximately 85–90% with lower energy depletion at end devices, whereas Random Forest and XGBoost provide reported accuracy ranges of approximately 90–95% and 92–95%, respectively, with substantially higher coordinator energy depletion. The results highlight the coordinator as a critical energy bottleneck in star-topology WSNs because traffic from multiple end devices converges at the coordinator. Based on the evaluated scenario, the study proposes a hybrid energy-aware detection strategy in which lightweight monitoring can be combined with selective use of computationally stronger ensemble classifiers. The study does not claim experimental validation against adaptive or adversarial LDoS attacks, and broader validation using independent scenarios, additional network sizes, and comprehensive classification metrics is identified as future work.