Temporal Convolutional Network Approach to Secure Open Charge Point Protocol (OCPP) in Electric Vehicle Charging

Securing data transactions across computer networks presents significant challenges, particularly concerning data privacy and cybersecurity threats. These issues are particularly critical in electric vehicle charging stations (EVCSs) due to the sensitive data involved, such as data breaches, unauthorized access, and privacy concerns. The primary challenge within EVCS architecture lies in defending against various cyberattacks. Several machine learning models, including convolutional neural networks, recurrent neural networks, and long short-term memory, have been employed to enhance EVCS security. However, these models often struggle to effectively handle the temporal dependencies, complexity, and scalability required for real-time threat detection. In this study, we propose a novel temporal convolutional network (TCN) to enhance EVCS security by rapidly detecting Open Charge Point Protocol (OCPP) attacks. Our model was trained and tested on the latest CICEVSE2024 dataset, which encompasses attacks targeting EVCS communication protocols, including OCPP and ISO15118, during charging and idle states. The proposed model achieved 100% accuracy in detecting binary labels and 5-class scenarios, and 93% accuracy in identifying 17-class attack types, surpassing existing approaches. Additionally, we conducted a comparative analysis with a hybrid model, demonstrating that the proposed TCN delivers superior performance with lower computational complexity, fewer parameters, a smaller model size, and shorter computation times. These features make the TCN model highly suitable for real-time EVCS applications. This work establishes a new benchmark for EVCS security, showcasing the potential of TCNs to deliver scalable and efficient solutions for safeguarding critical charging infrastructure against cyber threats.