Machine Learning-Based Intrusion Detection in Electric Vehicles with Hybrid Deep Neural Networks
Shakir Hussain, Dulam Devee Sivaprasad, G. Sravani, Thulasi A, Venkata Subramanian N. · 2025
Security problems, especially in the deployment of Internet of Things (IoT) devices, have grown in tandem with the paradigm shift towards integrating ordinary activities into the IoT. Thanks to developments in IoT technology, smart city initiatives like EVCSs—electric vehicle charging stations—have become quite popular. Intelligent features at these charging stations improve user convenience and provide operators more control. Having said that, they are vulnerable to cybersecurity risks due to their inherent Internet connectivity. Here, we lay forth a solid plan to beef up the intrusion detection systems (IDS) that protect EVCSs. To guarantee high-quality data for analysis, the preprocessing phase fixes issues including imbalance, missing values, outliers, and normalisation. The following state-of-the-art neural network methods are used for performance assessment: GCNN, Att-CNN, AGCNN, TransLST, and BiGRU. In terms of accuracy (98.05%), recall (96%), and F1 score (98.15%), the AGCNN model outperforms all of the others. The suggested AGCNN-based intrusion detection system (IDS) will improve the reliability and safety of EVCSs by reducing the impact of cyberattacks on daily life. The importance of implementing sophisticated machine learning models to guarantee the security and dependability of infrastructures that include the Internet of Things is emphasised in this study.