Performance Evaluation of Machine Learning Models for Intrusion Detection in Wireless Sensor Networks: A Case Study Using the WSN DS Dataset

Abstract

The security of Wireless Sensor Networks (WSNs) depends on effective Intrusion Detection Systems (IDSs), which are susceptible to Denial of Service (DoS) attacks. This research paper assesses the performance of three machine learning models, namely KNN, Logistic Regression, and Decision Tree (DT), using the WSN DS dataset for WSN intrusion detection. The dataset featured DoS attacks of several types: scheduling, blackhole, flooding, and gray hole attacks. Each model’s performance metrics were calculated and compared, including precision, recall, and F-1 score. Results showed that the DT model consistently outperformed the other models, demonstrating its effectiveness in accurately predicting different types of DoS attacks. The DT model exhibited superior performance with respect to macro-precision, macro-recall, and macro-F-1 score, achieving values of 0.98 each. In contrast, the logistic regression and kNN models yielded lower values of 0.98, 0.96, 0.97, and 0.87, 0.85, 0.86, respectively. These findings have significant implications for practitioners and researchers working on securing WSNs against DoS attacks and highlight the importance of using machine learning-based IDSs to detect and mitigate security threats in WSNs. This study enhances our knowledge of the detection of intrusions in WSNs. It offers guidance for creating strong security measures to ensure these networks’ dependable and secure functioning.