An Adaptive Spatio‐Temporal Stacked Ensemble Framework for Intelligent Anomaly Detection and Security Enhancement in Vehicular Ad Hoc Networks

ABSTRACT Vehicular Ad Hoc Networks (VANETs) play an essential role in intelligent transportation systems. However, the dynamic mobile nature of VANETs makes them vulnerable to a wide range of anomalous behaviors. To address these issues, we propose X‐ExTEN‐ID, an explainable stacked ensemble framework for adaptive and intelligent anomaly detection and vehicular security enhancement in VANET environments. The proposed framework involves spatiotemporal data analysis and multiple ensemble learning techniques combined through a stacked meta‐learning architecture to identify anomalous behaviors in real‐time VANET environments accurately. Instead of traditional single‐model approaches, our ensemble strategy combines multiple base learners to enhance detection accuracy and ensure protection against evolving attack strategies. To validate our suggested approach, we utilized the VeReMi dataset, which provides real‐world urban vehicular mobility tasks with labeled position falsification and other attack types. Experiment results demonstrated that our framework achieves notable scores in detection performance, with the highest recall rate of 98.9%, an F 1‐score of 99%, and a notably low False Positive Rate (FPR) of 1.3% and False Negative Rate of 0.9%, compared to existing machine learning models.