Attack-Label-Free Calibration, Temporal Splits, and Label-Budget Sensitivity in One-Class SCADA Anomaly Detection: Code and Processed Results
Year: 2026
Type: software
Abstract: Reproducibility release for the study “Label-Free Threshold Calibration and Label Efficiency in One-Class SCADA Anomaly Detection.” It contains the benchmark code, dataset configurations, retained per-seed scalar metrics, representative seed-42 score arrays, calibrated thresholds, final figures, the consolidated CSV result table, pinned environment requirements, and SHA-256 checksums. Raw third-party datasets are not redistributed; their original sources and expected file layout are documented in the archive. The benchmark evaluates labelled validation-F1 versus normal-only quantile calibration, event-level behavior, label-budget sensitivity, and chronological versus random splitting across BATADAL, Gas Pipeline, and HAI 22.04. Source repository: https://github.com/mharibat/scada-label-free-benchmark
Author: Haribat, Mohammad
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7204078341