Robustness of Anomaly Detection Models for Industrial Control Systems under Training-Time Data Contamination
2026-08-24 • Cryptography and Security
Cryptography and SecurityMachine Learning
AI summaryⓘ
The authors studied how well different machine learning models for detecting unusual activity in industrial control systems work when their training data gets messed up. They tested 11 detectors with three ways of adding bad or noisy data into the training sets, ranging from small to moderate amounts of contamination. Their findings show that some models, like PCA and SVM, handle corrupted training data better, while others, especially those relying on local data density, perform much worse. The authors highlight that good training data quality is very important for reliable anomaly detection in industrial systems.
Anomaly detectionIndustrial control systems (ICS)Machine learningTraining data contaminationOffline evaluationRobustnessPrincipal Component Analysis (PCA)Support Vector Machine (SVM)Local-density detectorsSecure Water Treatment (SWaT) dataset
Authors
Mustafa Umut Ozbek, Taiwo Ojo, Pooria Madani, Khalil El-Khatib, Li Yang
Abstract
Machine-learning-based anomaly detection is increasingly used in industrial control systems (ICS), yet most studies assume that detector training data is trustworthy. In practice, training data may be corrupted through compromised logs, labeling errors, manipulated historian records, or unsafe retraining processes. This paper evaluates the robustness of offline ICS anomaly-detection pipelines on the Secure Water Treatment (SWaT) benchmark under training-time contamination. We assess 11 heterogeneous anomaly detectors under three contamination strategies: random injection, similarity-targeted injection, and feature-noise injection. The first two insert attack samples into the nominal training pool, while the third adds bounded Gaussian noise to selected normal training samples. These attacks are contamination-based rather than gradient-driven poisoning methods. Contamination budgets from 1% to 10% are evaluated using clean validation and test sets under a unified offline protocol. The results show that robustness is strongly model-dependent and cannot be predicted from clean-data performance alone. Injection-based contamination causes the greatest degradation, particularly for local-density and distance-based detectors, whereas feature-noise contamination has a comparatively limited effect. PCA, SVM, HBOS, and IForest remain relatively stable, while the tuned neural detectors demonstrate intermediate robustness. Overall, the findings highlight the importance of training-data integrity in ML-enabled ICS monitoring, subject to the evaluated dataset, models, and threat assumptions.