New AI algorithm secures EV charging from cyberattacks with 2-stage correction

Researchers have developed a secure, two-stage AI algorithm to protect cloud-based electric vehicle charging from sensor data attacks. The method uses a self-learning Koopman operator to accurately estimate battery pack voltage even when adversaries corrupt measurement data during transmission. Tested with high-fidelity simulations, it reliably works across different battery configurations, ages, and attack types without needing excessive data or redundant sensors, ensuring optimal EV charging under compromised conditions.