Misbehavior Detection for Position Falsification Attacks in VANETs Using Ensemble Machine Learning

Cooperative Intelligent Transport System (CITS) is a modern approach to safety and efficient traffic operation over Vehicle Ad-hoc Networks (VANETs). This approach is based on vehicles communication each with other or infrastructure communication through wireless connections. The security of VANETs has been identified as a vital component in C-ITS, because there may some attacks in such network that can expose the safety of the driver and passengers or other road users. Intrusion Detection Systems (IDS) is a crucial instrument that assists to defend the vehicular network by detecting the misbehaving vehicles. In this paper, a new ML technique that capitalizes on three new features basically related to the sender location, which in turn improves the performances in position transgression attacks is developed. In the final, the output of each considered K-Nearest Neighbor (KNN), Random Forest (RF) and Decision Tree (DT) is given to Stacking method such as Generalized Linear Model (GLM) in order to bring superiority in the detection performance. An IDS is constructed allowing vehicles to detect misbehavior in a distributed way, while the detection mechanism is trained centrally. The findings show that the proposed system outperforms the existing techniques using classification with an accuracy of $\mathbf{9 9. 8 2 \%}$, precision of $\mathbf{9 7. 3 4 \%}$, recall of $\mathbf{9 7. 5 6 \%}$, and F1-score of $\mathbf{9 7. 2 8 \%}$.