A Robust Multi-Layer Stacking Model for Malicious Node Classification in Vehicular Networks

Abstract

Vehicular Ad Hoc Networks (VANETs) play a vital role in modern Intelligent Transportation Systems by enabling real-time communication between vehicles and infrastructure to enhance road safety and traffic efficiency. With the rapid advancement of 5G networks and the rise of connected and autonomous vehicles, VANETs have evolved into highly dynamic and data-intensive environments. Traditionally, security in these networks relied on rule-based mechanisms and signature-based Intrusion Detection Systems (IDS), which are no longer effective due to their inability to adapt to evolving attack patterns and large-scale data. The framework incorporates classifiers such as K-Nearest Neighbors (KNN), Support Vector Classifier (SVC), and Gaussian Naive Bayes (GNB), along with a hybrid ensemble model called Stacked Tao Tree (STT). Additionally, advanced data balancing methods, including Adaptive Synthetic Sampling (ADASYN) was employed to address class imbalance and improve learning performance. Experimental results demonstrate that the proposed STT model achieves exceptional accuracy, reaching 100.00% for denial-of-service attacks, 99.90% for sybil attack attempts, and 100.00% for blackhole attacks. The study provides a scalable and efficient solution for secure vehicular communication in dynamic network environments.