Enhanced Security Measures in Wireless Sensor Networks: Leveraging Random Forest and K-means Clustering for Node Replication Attack Detection

Abstract

During data transmission, a major security threat is posed by the node replication attacks in Wireless Sensor Networks (WSNs) where data integrity gets comprised. Initially, the projected work tells the vulnerabilities related to the node replication attacks where the sensor node's identity is replicated by the intruders leading to the network disruption. A novel strategy is implemented by integrating the unsupervised learning algorithm ‘K-means Clustering' with supervised learning algorithm ‘Random Forest Classifier' in this research work to find the node replication attacks effectively by these machine learning methods. The partitioning model of K-means Clustering will cluster the nodes by their similar behaviors. Along with K-means clustering, the ensemble learning approach, Random Forest is utilized to process the feature selection. The US Airforce LAN dataset is employed in this work and a reliable model for node replication attack detection is developed by Random Forest which categorizes the features as normal and intruder nodes after clustering. The accuracy of detection is evaluated by these methods after conducting several experiments on this dataset which measures the efficiency of false positives detection rate. The satisfactory execution results from the initial findings with significant improvements on conventional security methods. The improved level of endurance and precision is exhibited where risk mitigation is effectively achieved by this proposed work posed in WSN by node replication attacks.