Spanish Researchers Propose AI Agents to Secure EV Charging Networks

Researchers at the University of Malaga’s NICS lab have developed a multi-agent AI system designed to detect cyberattacks, fraud, and operational failures at electric-vehicle charging stations. The proposal places multiple AI agents at each station or network component.

Each agent assesses charger status, communications, and connected devices, then shares observations with neighboring agents through a central monitoring system.

The system operates under the Open Charge Point Protocol, which manages user authentication, electrical load, consumption monitoring, and diagnostics across charging networks. Current OCPP monitoring typically examines only network traffic or local events, limiting visibility across wider regions.

To address this gap, the architecture incorporates a consensus mechanism based on opinion dynamics, a mathematical model that mimics human social-network information exchange.

Agents gradually adjust assessments to reach collective agreement and reduce false positives. All agent transactions are recorded on a blockchain ledger that cannot be altered afterward, providing traceability and integrity for the monitoring process. The team tested the system in a simulated OCPP-compliant environment.

Agents encountered component failures, communication link errors, and scenarios requiring coordinated multi-part responses. The system identified both device-specific anomalies and behavioral patterns affecting multiple stations. The consensus mechanism improved diagnostic accuracy by cross-referencing observations from different agents.

Alcaraz is the lead author of the report published in the International Journal of Critical Infrastructure Protection.