Accuracy Evaluation of Anomaly Detection Models for Critical Infrastructure Considering False Alarms in Practical Deployment 


Vol. 15,  No. 3, pp. 238-246, Mar.  2026
https://doi.org/10.3745/TKIPS.2026.15.3.238


PDF
  Abstract

False positives in anomaly detection can lead to critical operational disruptions, such as the shutdown of power generators or water treatment facilities. Nevertheless, conventional accuracy evaluations often insufficiently account for false positives, allowing detection models with frequent false alarms to receive undeservedly high performance ratings. This study introduces a novel evaluation approach that incorporates false positive considerations alongside accuracy. The proposed approach identifies candidate thresholds that satisfy a predefined false positive rate and selects the threshold that maximizes accuracy for model evaluation. This procedure is essential, as anomaly detection models typically output probabilities or scores, making appropriate threshold determination crucial. Experimental validation using two datasets, three accuracy metrics, and nine detection models demonstrates that the proposed approach yields performance assessments that differ substantially from those obtained through conventional evaluation approaches.

  Statistics
  Cite this article

[IEEE Style]

W. Hwang, J. Yun, Y. Kim, J. Kim, B. Min, "Accuracy Evaluation of Anomaly Detection Models for Critical Infrastructure Considering False Alarms in Practical Deployment," The Transactions of the Korea Information Processing Society, vol. 15, no. 3, pp. 238-246, 2026. DOI: https://doi.org/10.3745/TKIPS.2026.15.3.238.

[ACM Style]

Won-Seok Hwang, Jeong-Han Yun, Yesol Kim, Jonguk Kim, and Byung-Gil Min. 2026. Accuracy Evaluation of Anomaly Detection Models for Critical Infrastructure Considering False Alarms in Practical Deployment. The Transactions of the Korea Information Processing Society, 15, 3, (2026), 238-246. DOI: https://doi.org/10.3745/TKIPS.2026.15.3.238.

Forms

Search
(IN TITLE, AUTHOR, ABSTRACT,KEYWORDS)

Advanced Search ▼

POPULAR KEYWORDS
(TOP 10 KEYWORDS)

Recent Publications
(LAST 3 YEARS)

Old Journals

Indexing

All publications of TKIPS is indexed in DOI, EBSCO, Google Scholar, Crossref, and CrossCheck.

TKIPS is also selected as the Journal for Accreditation by NRF (National Research Foundation of Korea).

Related Journals