BurST-ADMA: Towards an Australian Dataset for Misbehaviour Detection in the Internet of Vehicles
Mohamed Ahzam Amanullah, Mohan Baruwal Chhetri, Seng W. Loke, Robin Doss
Abstract
The Internet of Vehicles (IoV) is a major contributor to Cooperative Intelligent Transportation Systems (C-ITS) as it enables intercommunication between the connected vehicles and the supporting infrastructure. Basic Safety Messages (BSMs) are a special type of messages used by connected vehicles to communicate kinematic information including position, speed, acceleration, and heading, with other nodes in C-ITS. However, these BSMs are susceptible to false data injection attacks that can disrupt the normal (cooperative) functioning of the C-ITS leading to collisions or traffic jams. While there are several data-centric misbehaviour detection mechanisms proposed in the literature, most of them have been evaluated using either the Vehicular Reference Misbehaviour (VeReMi) dataset or the VeReMi Extension dataset. Both datasets are created for the Luxembourg SUMO Traffic (LuST) scenario, making them appropriate for evaluating misbehaviour detection mechanisms in an European C-ITS context. However, there are no publicly available misbehaviour datasets that are representative of a non-European C-ITS context. To address this shortcoming, we propose a new scenario called the Burwood SUMO Traffic (BurST) scenario, which is modelled on the suburb of Burwood in Melbourne, Victoria, Australia. We also introduce the preliminary misbehaviour dataset for the BurST Scenario – BurST-Australian Dataset for Misbehaviour Analysis (BurST-ADMA) – which includes false data injections in the position, speed, and heading of the BSMs. To the best of our knowledge, this is the first public extensible Australian dataset for C-ITS misbehaviour detection studies. Finally, we evaluate the BurST-ADMA dataset with some commonly used machine learning techniques discussed in the literature.