Reinforcement Learning-Based Intrusion Detection for Electric Vehicle Charging Stations

Electric Vehicle Charging Stations (EVCS) face an evolving threat landscape where cyber-attacks can compromise operational integrity and user data. Traditional Intrusion Detection Systems (IDS) often rely on static modeling approaches, which fail to adapt to novel or zero-day attack vectors. In this paper, we propose a novel Reinforcement Learning (RL)-based IDS framework using Deep Q-Network (DQN) to detect port scans, flood attacks, and advanced threats in EVCS environments. The system extracts discriminative features from network traffic and adaptively updates detection policies based on environment feedback. Evaluated on the CICEVSE2024 dataset, the proposed approach demonstrates significant improvements in detection accuracy, reduced false positive rates, and efficient computational overhead compared to conventional baseline IDS solutions. Our results suggest that the adaptive decision-making capability of RL can enhance cybersecurity defenses in EVCS, offering a more robust and future-proof solution for modern automotive power infrastructures.