Detect and Mitigate Denial of Service (DoS) Attacks Using Lightweight ML

Detect and Mitigate Denial of Service (DoS) Attacks Using Lightweight ML

Authors- Ms. Nidhi Singh

Abstract-Wireless Sensor Networks are increasingly deployed in critical applications, yet their resource-constrained nature makes them vulnerable to Denial of Service (DoS) attacks. Traditional security mechanisms often fail due to high compu- tational overhead, rendering them impractical for WSNs. This paper proposes DoSGuard, a lightweight MLframework designed to detect and mitigate DoS attacks in WSNs. Leveraging a hybrid simulation and training pipeline, we integrate feature engineering with the CatBoost algorithm to achieve high detection accuracy while maintaining low resource demands. Our approach simulates a 500-node WSN, engineers features such as packet rate changes and energy levels, and employs CatBoost for classification. Evaluation results demonstrate an accuracy of over 95%, with precision and recall exceeding 94%, validated through extensive visualizations including ROC curves and confusion matrices. The framework further includes a mitigation strategy that filters malicious traffic, reducing attack impact by up to 80%. This work offers a scalable, efficient solution for securing WSNs against DoS threats, suitable for real-world deployment.

DOI: /10.61463/ijset.vol.13.issue3.104