Intrusion Detection in Wireless Sensor Networks using SMOTE Tomek Link sampling technique

Wireless Sensor Networks (WSNs) are an essential component of cyber-physical systems, characterized by the integration of stationary and mobile sensors that collaboratively capture and transmit environmental data. These sensors utilize self-organization and multi-hop communication mechanisms, which enable them to efficiently gather and process information from their surroundings. Despite the advantages provided by WSNs, they are vulnerable to various attacks that can severely compromise their functionality, leading to a pressing need for effective intrusion detection systems. Traditional intrusion detection methods in WSNs face significant challenges, including low detection rates, imbalanced & complex higher dimensional datasets and false alarms. To address the aforementioned challenges, an innovative intrusion detection approach is proposed, which integrates Machine Learning (ML) techniques with the Synthetic Minority Oversampling Technique Tomek Link (SMOTETomekLink) algorithm. This hybrid methodology aims to synthesize minority instances within the dataset while removing Tomek links, thereby creating a balanced dataset that enhances the accuracy of intrusion detection systems tailored for WSNs. The application of feature scaling through standardization is another critical element of the proposed model. Furthermore, implementing the SMOTETomek resampling technique is essential for counteracting imbalances in WSN datasets, which addresses the persistent issues of overfitting and underfitting often observed in machine learning applications. A comprehensive evaluation of the proposed intrusion detection model was conducted using the Wireless Sensor Network Dataset (WSN-DS), which comprises 374,661 records. ​The findings achieved a remarkable accuracy rate of 99.78% in binary classification scenarios.​ This high level of accuracy demonstrates the efficacy and superiority of the developed system within the context of WSN intrusion detection.