EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Scalable Hybrid Online Machine Learning for Real-Time Intrusion Detection in Electric Vehicle Charging Systems

Josiah Kipruto, Bernardi Pranggono, Raj Mani Shukla

IEEE Access · 2026 · 14 · 12707-12716

DOI: 10.1109/access.2026.3656666 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IEEE Access, 2026. Verfasst von Josiah Kipruto und 2 weiteren Autor:innen. Die Arbeit wurde bisher 2-mal zitiert (Crossref) und nennt 24 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Electric Vehicle Charging Systems (EVCS) are increasingly integrated into smart grids, but their openness and connectivity expose them to diverse cyber threats. Intrusion Detection Systems (IDS) are essential for securing these infrastructures; however, conventional offline models struggle with real-time data streams, evolving attack patterns, and concept drift. This study investigates online machine learning for real-time IDS in EVCS. Two approaches are evaluated: a baseline Adaptive Random Forest (ARF) with ADWIN drift detection, and a scalable hybrid ensemble integrating ARF with Naïve Bayes, k-Nearest Neighbors, and Logistic Regression under a meta-learning framework. Experimental results show that while the baseline ARF achieves competitive performance, it incurs high computational cost (267.98 minutes for the 15-class task). In contrast, the hybrid model reduces runtime to 81.65 minutes while improving accuracy (0.8582 vs. 0.8142), precision (0.8796 vs. 0.8341), recall (0.8582 vs. 0.8142), and F1-score (0.8344 vs. 0.8002). Notably, in binary classification, the hybrid ensemble achieves near-perfect detection (F1=1.0000) while maintaining stability under drift-free conditions. These findings demonstrate that combining diverse classifiers enhances drift resilience, improves computational efficiency, and strengthens real-time detection, offering a scalable IDS solution tailored to EVCS and advancing the security and reliability of modern smart grids.

Zitieren

APA

Kipruto, J., Pranggono, B., & Shukla, R. M. (2026). Scalable Hybrid Online Machine Learning for Real-Time Intrusion Detection in Electric Vehicle Charging Systems. IEEE Access, 14, 12707-12716. https://doi.org/10.1109/access.2026.3656666

Vancouver

Kipruto J, Pranggono B, Shukla RM. Scalable Hybrid Online Machine Learning for Real-Time Intrusion Detection in Electric Vehicle Charging Systems. IEEE Access. 2026;14:12707-12716. doi:10.1109/access.2026.3656666

BibTeX
@article{Kipruto2026,
  author = {Kipruto, Josiah and Pranggono, Bernardi and Shukla, Raj Mani},
  title = {Scalable Hybrid Online Machine Learning for Real-Time Intrusion Detection in Electric Vehicle Charging Systems},
  journal = {IEEE Access},
  year = {2026},
  volume = {14},
  pages = {12707--12716},
  doi = {10.1109/access.2026.3656666}
}
Kostenlos starten →