kumoh national institute of technology
Networked Systems Lab.

Goodness Oluchi Anyanwu, Cosmas Ifeanyi Nwakanma, Jae-Min Lee, Dong-Seong Kim, "Ensemble Random Forest-based Basic Safety Messages Falsification Detection in Internet of Vehicles.", IEEE Access (Minor Edits- Resubmit)
By : Goodness
Date : 2022-05-16
Views : 74

Misbehavior control and detection is a critical advancement to guarantee that shared data is certified on the Internet of Vehicles (IoV) as information about vehicles and their movement states is exchanged in real-time. However, IoV is susceptible to misleading safety messages that could create life-threatening situations. Thus, detecting nodes that are propagating inaccurate information is a requirement for the successful deployment of IoV services. As such, this paper proposes a Falsification Detection Scheme (FDS) based on the concept of the Ensemble Random Forest (Ens. RF) Algorithm. The Randomized Search Cross-Validation technique was used to construct the proposed novel Ens. RF. The evaluation was performed on the Vehicular Reference Misbehaviour (VeReMi) dataset, which was developed to tackle and encourage misbehaviour research in IoV. Also, four other Machine Learning (ML) algorithms were investigated to evaluate the capability of the proposed Ens. RF algorithm, which had the best performance using the preprocessed dataset with five falsifications and one benign category. The performance metrics considered are computing efficiency, validation accuracy for overall attack classification, precision, recall, and F1 scores. For validation, the performance of the proposed Ens. RF was further compared with recent works. The result shows that the proposed Ens. RF outperformed state-of-the-art algorithms implemented in this work and related works with an overall validation accuracy of 99.60\% and a negligible $604$ mislabeled points red out of 153730 points