Multi-Layer Anomaly Detection Framework for Industrial Control Systems: A Comprehensive Comparative Analysis of Unsupervised Machine Learning Algorithms

Key Points

• To develop and evaluate a multi-layer anomaly detection framework for Industrial Control Systems using unsupervised machine learning algorithms.
• Created a multi-layer framework encompassing IoT, OT (Operational Technology), Linux, Windows, and network layers.
• Evaluated various unsupervised machine learning algorithms to detect anomalies effectively.
• Assessed performance against realistic false positive rate constraints.
• Identified effective algorithms for anomaly detection across different layers of Industrial Control Systems.
• Demonstrated the importance of layer-specific strategies for minimizing false positive rates.