Enhancing Security in Wireless Sensor Networks: A Machine Learning-based DoS Attack Detection
Ghadeer Al Sukkar, Saleh H. Al-Sharaeh · Engineering Technology & Applied Science Research · 2025
The Internet of Things (IoT) is based on Wireless Sensor Networks (WSNs), which are essential for many applications. Denial of Service (DoS) attacks are a major risk for WSNs due to their open architecture and limited resources. This paper investigates how different Machine Learning (ML) methods can be used to identify DoS attacks and mitigate their effects. The predictions from several models were combined using the ensemble method to increase overall accuracy, while explainable Artificial Intelligence (AI) techniques were also deployed to enhance transparency and understanding. To compare the performance of both hard and soft ensemble methods, the WSN Dataset (WSN-DS) and the WSN Blackhole, Flooding, and Selective Forwarding (WSN-BFSF) dataset were utilized. The ensemble techniques aggregated predictions from multiple models to improve overall accuracy, while both showed high accuracy for both datasets. With an accuracy of 98.12%, the soft ensemble technique slightly outperformed the hard ensemble technique for the WSN-DS dataset, which had an accuracy of 97.97%. For the WSN-BFSF dataset, the hard ensemble technique achieved an accuracy of 99.967%, while the soft ensemble technique achieved an excellent accuracy of 100%.