Advanced Temporal Convolutional Network Framework for Intrusion Detection in Electric Vehicle Charging Stations

Abstract

Electric Vehicle Charging Station (EVCS) systems have become increasingly critical to the energy and transportation sectors. The detection of various attacks in EVCS, including data interception in the Open Charge Point Protocol (OCPP), poses substantial cybersecurity challenges that existing deep learning methods struggle to address effectively. This work investigates the impact of 16 types of attacks on EVCS, such as denial-of-service (DoS), reconnaissance, cryptojacking, and backdoor attacks. To address these threats, we propose an innovative model designed to identify diverse cyber threats targeting EVCS. The proposed Temporal Convolutional Network (TCN)-based Intrusion Detection System (IDS) architecture integrates four key innovations: multi-receptive fields, a gating mechanism, iterative dilation, and a self-attention mechanism combined with a Squeeze-and-Excitation (SE) block to recalibrate feature responses by explicitly modeling interactions between different channels. The proposed model effectively processes multiple temporal scales, regulates the flow of information, adapts to varying time steps, and focuses on essential components of time-series data. Experimental evaluations demonstrate that the proposed model outperforms state-of-the-art methods in terms of accuracy and detection rates across all 16 attack types in the CICEVSE2024 dataset, which comprises extensive attack vectors and variants associated with the OCPP. The proposed approach achieves higher accuracy compared to other TCN variants and exhibits high resilience against complex attacks.