Abstract:
Vehicular Ad-hoc Networks (VANETs) are a cornerstone of Intelligent Transportation Systems
(ITS), enabling vehicle-to-vehicle and vehicle-to-infrastructure communication to improve road
safety and optimize traffic flow. Despite their critical importance, VANETs face numerous
security threats, particularly from insider attacks that compromise the accuracy and reliability of
transmitted data. Among these threats, position falsification where malicious actors disseminate
incorrect Basic Safety Messages (BSMs) poses significant risks, including traffic disruptions and
potential accidents.
This study explores a data-centric machine learning approach to detect and mitigate position
falsification attacks within VANETs. Specifically, we evaluated three machine learning algorithms
Decision Tree, Random Forest, and K-Nearest Neighbors (KNN) each integrated with bagging
techniques to enhance classification performance. Among these, KNN with bagging techniques
achieved superior detection rates compared to the others.
The detection model was trained and validated using the Vehicular Reference Misbehavior Dataset
(VeReMi), which encompasses five distinct types of position falsification attacks. Our proposed
system operates at the roadside unit (RSU) level, analyzing sequential BSM transmissions to
identify malicious behavior while minimizing computational overhead for individual vehicles. The
KNN with bagging model demonstrated exceptional detection capabilities, achieving a 100%
detection rate for Attack 1, 99.9% for Attack 2 and Attack 4, 99% for Attack 8, and 97.9% for
Attack 16. The system maintained high performance across all key evaluation metrics, including
accuracy, precision, recall, and F1 score.
By leveraging the RSU-level architecture, this approach not only ensures efficient detection and
mitigation of malicious nodes but also offers a scalable solution for enhancing VANET security.
The findings highlight the effectiveness of KNN with bagging techniques in providing robust
protection against position falsification attacks, thereby contributing to the overall safety and
reliability of VANET communication systems.