Ensemble Classifier-based Intrusion Detection System for WSNs: Optimizing Security Against Evolving Attacks

This paper presents an efficient Intrusion Detection System specified to the requirements of WSNs, targeting blackholes, floods, and scheduling-related intrusion attacks. The increasing vulnerability of WSNs to evolving attack patterns and their constrained resources make traditional security solutions, such as cryptography and trust-based systems, inadequate. The most widely used security technologies employ cryptography and trust-based systems have built-in shortcomings at protecting WSNs against evolving attack patterns and constricted resources. The proposed IDS will circumvent these challenges because it uses ensemble classifiers, namely REP Tree, Random Forest, and JRip, that demonstrate significant performance in the heavily constrained environment and unbalanced datasets. The results show that the system outperforms traditional IDS methods, with reduced resource consumption and a low false positive rate. Using the WSN-DS benchmark dataset, the system managed an average accuracy of 98.3% and a precision of 97.1%. This system is optimal for use in actual WSN installations because of its better performance than traditional IDS methods and small consumption of the resources.