Anomaly Detection in Cyber-physical Systems: A SWAT Data-Based Mathematicalmodelingapproach
Vìsnik Nacìonalʹnogo unìversitetu "Lʹvìvsʹka polìtehnìka". Serìâ Ìnformacìjnì sistemi ta merežì · 22.06.2026 · 19 · 76-87
DOI: 10.23939/sisn2026.19.076 ↗Kurz gefasst
Fachartikel in der Zeitschrift Vìsnik Nacìonalʹnogo unìversitetu "Lʹvìvsʹka polìtehnìka". Serìâ Ìnformacìjnì sistemi ta merežì, online erschienen am 17.05.2026. Verfasst von Nazar Ilchuk und 1 weiteren Autor:innen. Die Arbeit nennt 32 Literaturangaben.
Abstract (Originalsprache)
This work presents statistical anomaly detectionmethods for cyber-physical systems using the Secure Water Treatment dataset, a scaled down version of a real-world industrial water treatment plant. We consider a critical sensor signal, the water level indicator LIT301, and set up two corresponding models for detecting cyber-attacks - a sliding window Z-score outlier detection, and an autoregressive integrated moving average time-series forecasting model of order (3,0,5). The sliding window Z-score model checks the deviation of LIT301 readings from a local mean, flagging statistically meaningful anomalies. The ARIMA(3,0,5) model first is trained on the normal dynamics of LIT301 and then forecast errors (residuals) are flagging them as anomalies. We make an assessment of these models against three known cyber-attack scenarios provided in the SWaT dataset. First attack is the single-sensor spoofing attack on LIT301 sensor directly. Second and third attacks are held against two other detectors that indirectly affect LIT301, a multi-point coordinated attack on a valve (MV201) and a pump (P101), and an actuators attack turning off pump P301 We show in this work that both models successfully detect the anomalous behavior. We present mathematical formulations for the Z-score and ARIMA residuals, along with quantitative results and time-stamped detection of each attack. Both ARIMA and Z-score sliding window models are shown to detect all three attacks with high confidence. Here we discuss the performance of both methods, including detection delay and sensitivity, and give insight into their false-positive characteristics. Altogether, the obtained results show that simple statistical techniques can effectively detect complex cyber-attacks on CPS offering lightweight solutions for CPS anomaly detection.
Zitieren
APA
Ilchuk, N., & Veres, O. (2026). Anomaly Detection in Cyber-physical Systems: A SWAT Data-Based Mathematicalmodelingapproach. Vìsnik Nacìonalʹnogo unìversitetu "Lʹvìvsʹka polìtehnìka". Serìâ Ìnformacìjnì sistemi ta merežì, 19, 76-87. https://doi.org/10.23939/sisn2026.19.076
Vancouver
Ilchuk N, Veres O. Anomaly Detection in Cyber-physical Systems: A SWAT Data-Based Mathematicalmodelingapproach. Vìsnik Nacìonalʹnogo unìversitetu "Lʹvìvsʹka polìtehnìka". Serìâ Ìnformacìjnì sistemi ta merežì. 2026;19:76-87. doi:10.23939/sisn2026.19.076
BibTeX
@article{Ilchuk2026,
author = {Ilchuk, Nazar and Veres, Oleh},
title = {Anomaly Detection in Cyber-physical Systems: A SWAT Data-Based Mathematicalmodelingapproach},
journal = {Vìsnik Nacìonalʹnogo unìversitetu "Lʹvìvsʹka polìtehnìka". Serìâ Ìnformacìjnì sistemi ta merežì},
year = {2026},
volume = {19},
pages = {76--87},
doi = {10.23939/sisn2026.19.076}
}
EUVIMED