Detection Of Blackhole And Wormhole Attacks In Ad Hoc Networks Using Ensemble Learning

Abstract

Mobile Ad Hoc Networks (MANETs) are highly dynamic and infrastructure-less wireless networks that are vulnerable to various routing attacks due to their open medium, decentralized control, and rapidly changing topology. Among these, blackhole and wormhole attacks are considered highly destructive, as they disrupt routing mechanisms by either dropping packets maliciously or creating unauthorized tunnels to manipulate network traffic. Such attacks significantly degrade packet delivery ratio, increase end-to-end delay, and compromise overall network reliability. This research proposes an ensemble learning-based framework for the detection of blackhole and wormhole attacks in MANET environments. The proposed model integrates multiple supervised machine learning classifiers, including Decision Tree, Random Forest, and Gradient Boosting, to enhance detection accuracy and robustness. Network performance parameters such as packet delivery ratio, routing overhead, end-to-end delay, hop count variation, and sequence number anomalies are extracted as key features for classification. The ensemble approach combines the strengths of individual learners using majority voting and weighted aggregation strategies to minimize false positives and improve generalization.