ENHANCING SECURITY OF PROSUMER EV CHARGING STATIONS USING EDGE-ASSISTED FEDERATED LEARNING

Abstract

The rapid growth of electric vehicles (EVs) and smart charging infrastructures has increased the importance of securing prosumer-based EV charging stations against sophisticated cyber threats. Traditional centralized security mechanisms often face challenges such as privacy leakage, high communication overhead, and poor adaptability to heterogeneous network traffic data. To address these limitations, this work proposes an intelligent cybersecurity framework titled “Enhancing Security of Prosumer EV Charging Stations Using Edge-Assisted Federated Learning.” The proposed approach integrates edge computing with federated learning to enable decentralized and privacypreserving cyber-attack detection across distributed EV charging environments. Feature selection using Pearson Correlation Coefficient (PCC) is employed to improve detection accuracy and reduce computational complexity. Furthermore, knowledge distillation and prototype aggregation techniques are incorporated to effectively handle nonindependent and identically distributed (non-IID) network traffic data among prosumers. Each dedicated local edge server performs local training and collaboratively contributes to global model optimization without sharing raw data. After detecting malicious activities, a rule-based intervention mechanism is triggered to mitigate cyber threats in real time. Experimental evaluations conducted on benchmark datasets such as NSL-KDD, UNSW-NB15, and IoTID20 demonstrate that the proposed framework achieves superior detection accuracy, improved overall detection correctness, and reduced false alarm rates compared with existing federated learning methods. The proposed system provides a scalable, secure, and privacy-aware solution for next-generation intelligent EV charging infrastructures.