Electric Vehicle Charging Station (EVCS) security is an emerging concern in our interconnected world today owing to the growing complexity and rate of cyber threats. Traditional Intrusion Detection Systems (IDS) for EV charging infrastructure have challenges in detecting novel or unexpected attacks due to their limitation of predetermined signatures as well as other limited detection capabilities. The aim of this study is towards the development of a deep learning based security solution for electric vehicle charging which detects anomalies. The Host Events subdirectory of the CICEVSE2024 dataset for electric vehicle charging infrastructure security was adopted for developing the solution. After preprocessing, the distribution of the data was optimized by applying the Standard Scaling, Log Transform and Normalization methods. Tuning of hyperparameters such as number of epochs as and batch size were carried out to develop three distinct DL models for specific target variables of State, Scenario and Label. The Deep Learning solution was thereafter tested on the bases of performance metrics such as Accuracy, F1-Score, Precision, Recall as well as ROC-AUC values. The system proved to be of excellent performance across the three target variables. Comparative analysis of the various model performances showed that the Label target variable produced the most optimal results of 0.9960, 0.9957, 0.9946 and 0.9967 respectively for accuracy, F1-score, precision and recall respectively.
Moses (Thu,) studied this question.