Optimization Accuracy of Blackhole Attacks in Routing Protocol using Machine Learning

Abstract

Blackhole attacks pose a significant threat to the reliability and stability of wireless ad hoc and sensor networks, as malicious nodes intentionally drop data packets after falsely advertising optimal routes. Such attacks lead to degradation in network performance by reducing the packet delivery ratio, increasing end-to-end delay, and causing severe routing disruptions. Traditional defense techniques rely on threshold-based or heuristic mechanisms, which often fail to adapt to dynamic network conditions and result in low detection accuracy. To address these limitations, machine learning (ML)-based approaches have emerged as an effective solution for detecting and mitigating Blackhole attacks. This paper proposes an optimized ML framework aimed at enhancing the accuracy of identifying Blackhole nodes within routing protocols such as AODV, DSR, and DSDV. The model analyzes essential network behavior features including abnormal sequence numbers, packet drop rate, routing overhead, and transmission patterns to classify nodes as benign or malicious. Various ML classifiers—such as Random Forest, SVM, Logistic Regression, KNN, and ensemble learning—are evaluated using performance metrics like accuracy, precision, recall, F1-score, and ROC curves.