Effective Long Short-Term Memory based-Driver Identification in ITS

Human driving patterns are individualized and distinct, and the vehicle fingerprint of drivers could aid in automatically identifying different driving behaviors and be used in fields such as anti-theft systems. The introduction of in-vehicle networking systems, along with cutting-edge sensors and communication technologies have enabled the collection of massive volumes of nearly real-time data about vehicles and drivers, thereby expanding future possibilities. This research work presents a Deep learning-based LSTM strategy to classify the drivers based on the driving characteristics of the drivers collected by OBD-II protocols from vehicles’ ECUs. In contrast to previous research, a cross-validation technique is employed to obtain reliable results when applied to actual data. An accuracy of above 99 percent has been obtained by assessing a dataset comprised of the data gathered from actual vehicles. The work demonstrated a better performance in terms of evaluation metrics (Accuracy, Precision, Recall, and F-1 Score) when compared with other Machine Learning (ML) and Deep Learning (DL) models.