This article presents a novel intrusion detection system (IDS) for IoT-based electric vehicle charging stations (EVCS), integrating convolutional neural networks (CNN), long short-term memory (LSTM), and gated recurrent units (GRU) to enhance cybersecurity. The proposed framework leverages a comprehensive real-world dataset, Edge-IIoTset, to address the unique challenges of IoT-based EVCS. The model achieves 100% accuracy in binary classification, 97.44% in six-class classification, and 96.90% in fifteen-class classification, demonstrating its effectiveness in detecting intrusions. The ensemble architecture combines the strengths of CNN, LSTM, and GRU to provide a robust and adaptive IDS for IoT environments. The model is implemented using Python, TensorFlow, and Keras, and is available on GitHub. The study highlights the importance of advanced neural network models in enhancing cybersecurity for IoT-based EVCS and sets a new benchmark in the field. The results underscore the potential of deep learning in improving intrusion detection accuracy and resilience in IoT systems.This article presents a novel intrusion detection system (IDS) for IoT-based electric vehicle charging stations (EVCS), integrating convolutional neural networks (CNN), long short-term memory (LSTM), and gated recurrent units (GRU) to enhance cybersecurity. The proposed framework leverages a comprehensive real-world dataset, Edge-IIoTset, to address the unique challenges of IoT-based EVCS. The model achieves 100% accuracy in binary classification, 97.44% in six-class classification, and 96.90% in fifteen-class classification, demonstrating its effectiveness in detecting intrusions. The ensemble architecture combines the strengths of CNN, LSTM, and GRU to provide a robust and adaptive IDS for IoT environments. The model is implemented using Python, TensorFlow, and Keras, and is available on GitHub. The study highlights the importance of advanced neural network models in enhancing cybersecurity for IoT-based EVCS and sets a new benchmark in the field. The results underscore the potential of deep learning in improving intrusion detection accuracy and resilience in IoT systems.