ADAPTIVE INTRUSION DETECTION SYSTEMS FOR WIRELESS SENSOR NETWORKS: A HYBRID MACHINE LEARNING APPROACH

Nowadays, Wireless Sensor Networks are growing rapidly and a considerable portion of telecommunication. Wireless sensor networks are made up of tiny, low powered sensor nodes, which perceive information from their surroundings., analyse them and connect to the sink nodes wirelessly, have limitations and security threats due to their low memory and computational power. As the network is growing rapidly, it is critical to detect malicious activities in network. For safeguarding the sensitive data from vulnerabilities, Intrusion Detection System (IDS) emerges as a crucial tool. Various studies focus on the improved accuracy of the intrusion detection model and a reduced false alarm rate. Due to this, the elimination of repeating features is demonstrated. This research presents a hybrid classification approach for identifying malicious network packets using a simulated network environment. Our hybrid model uses an intelligent feature selection process that improvise the utility of dataset by preprocessing the data and tuning the feature set by including only relevant feature in dataset. After removing the redundant features, a subset of 41 is selected and narrowed down to 20 pivotal features using Neighbourhood Component Analysis (NCA) for effective and efficient analysis. The study involves a comprehensive comparison of K-Nearest Neighbors, Decision Tree, and Ensemble classifiers, and introduces a hybrid method that integrates these classifiers on the basis of voting mechanism to enhance detection accuracy and robustness. Performance is evaluated using key metrics such as accuracy, precision, recall, F-measure, detection rate, and false alarm rate. Experimental results demonstrate that the hybrid model significantly outperforms individual classifiers, achieving higher detection rates and lower false alarm rates. This study underscores the efficacy of ensemble techniques and hybrid models in cybersecurity applications, providing a robust framework for intrusion detection. The implications of this research are profound, offering an impressive 99.24% of accuracy rate with very low False Alarm Rate. By significantly boosting the IDS accuracy and minimizing the false alarm rate, This study advances WSN’s security posture against emerging cyber threats.