Romanian Cyber Security Journal / Spring 2026, No. 1, Vol. 8

Intrusion Detection in Intelligent Energy Prediction Systems for Non-Residential Buildings

Sorin SOVIANY, Eleonora TUDORA, Ovidiu BICA, Cristina-Gabriela GHEORGHE

---

Abstract

This paper addresses the cybersecurity of intelligent energy prediction systems in non-residential buildings by proposing a formally structured Multi-Layer Correlated Intrusion Detection System (MLC-IDS) aligned with the PRECONERG security-by-design architecture. The main contribution consists of extending traditional IoT intrusion detection by integrating AI pipeline integrity protection and cross-layer alert correlation within a unified detection framework. The proposed methodology introduces resource-aware edge-level anomaly detection, protocolaware behavioral modeling at the gateway level, hybrid signature–/Machine Learning-based network intrusion detection with adaptive fusion, and data poisoning and concept drift detection embedded directly in the prediction lifecycle. A formal alert aggregation model is defined to support the trade-off between detection accuracy and operational overhead. An experimental validation framework with Machine Learning models is proposed to demonstrate measurable improvements in detection reliability and prediction robustness under adversarial conditions.

Keywords

Intrusion Detection Systems, Multi-Layer Security Architecture, Non-Residential Smart Buildings, Intelligent Energy Consumption Prediction, IoT Security