Prospective Evaluation of Machine-Learning Intrusion Detection for IEC-104 Microgrid Communications
Journal of Intelligent Decision Making and Information Science · 09.09.2026 · 3 (18s) · 188-213
DOI: 10.17762/jidmis.v3.5724 ↗Kurz gefasst
Fachartikel in der Zeitschrift Journal of Intelligent Decision Making and Information Science, online erschienen am 09.09.2026. Verfasst von Mohamed Hachimi. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
Intrusion Intrusion detection in IEC 60870-5-104 networks must distinguish malicious activity from legitimate variations in network behavior. This study prospectively evaluates previously developed machine-learning pipelines on newly collected laboratory sessions and examines their sensitivity to the normalization of TCP connection-state features. Three family-held-out variants were assessed: a network-window baseline, a state-aware design incorporating established and closing TCP connection counts, and a normalization-sensitivity variant using the same state-aware features.The prospective dataset comprised 60 sessions: 30 benign sessions and 10 sessions each of port reconnaissance, targeted SYN pressure, and established-connection holding. Twelve benign sessions included an additional passive client. Evaluation used non-overlapping one-second windows, disjoint benign test sessions, a historically fixed threshold of 0.5, and stratified session-level bootstrap resampling.The state-aware V2 instances achieved F1 scores of 1.0000 for port reconnaissance, 1.0000 for SYN pressure, and 0.9915 for connection holding, with an unweighted mean of 0.997151. In contrast, the A3-held-out network-only and normalization-sensitivity instances failed to detect eligible connection-holding windows. These findings demonstrate strong prospective performance for the complete state-aware pipeline under the evaluated laboratory conditions, while revealing substantial dependence on preprocessing. They do not establish robustness across alternative normalization choices, benign-concurrency levels, attack intensities, device implementations, or operational deployments.
Zitieren
APA
Hachimi, M. (2026). Prospective Evaluation of Machine-Learning Intrusion Detection for IEC-104 Microgrid Communications. Journal of Intelligent Decision Making and Information Science, 3(18s), 188-213. https://doi.org/10.17762/jidmis.v3.5724
Vancouver
Hachimi M. Prospective Evaluation of Machine-Learning Intrusion Detection for IEC-104 Microgrid Communications. Journal of Intelligent Decision Making and Information Science. 2026;3(18s):188-213. doi:10.17762/jidmis.v3.5724
BibTeX
@article{Hachimi2026,
author = {Hachimi, Mohamed},
title = {Prospective Evaluation of Machine-Learning Intrusion Detection for IEC-104 Microgrid Communications},
journal = {Journal of Intelligent Decision Making and Information Science},
year = {2026},
volume = {3},
number = {18s},
pages = {188--213},
doi = {10.17762/jidmis.v3.5724}
}
EUVIMED