VAE-XGBoost: a hybrid intrusion detection system for next generation EV charging networks

The widespread adoption of electric vehicles (EVs) is crucial for reducing greenhouse gas emissions, yet it exposes charging infrastructure to sophisticated cyber threats. In the next generation EV charging networks, ultra-low latency, massive device connectivity, and AI-driven automation are key technologies. Traditional intrusion detection systems (IDS) struggle with class imbalance, overfitting, and real-time anomaly detection. These limitations threaten user privacy, service continuity, and grid stability. This study proposes a Variational Auto Encoder (VAE) based XGBoost, a hybrid IDS that leverages VAE for feature extraction and handling imbalanced attack data, while XGBoost improves classification accuracy. Evaluated on the CICEVSE2024 dataset, the model surpasses traditional methods like K-Nearest Neighbors and Random Forest, achieving an accuracy of 88.75%, a precision of 88.73%, a recall of 88.75%, and an F1-score of 88.67%. By integrating AI-driven anomaly detection into 6G-enabled smart grids, VAE-XGBoost enhances the cybersecurity resilience of EV charging infrastructure, ensuring scalability, adaptability to novel threats, and real-time mitigation for sustainable and secure transportation networks.