Rectilinear Hybrid Belief Classifier for Cyber-Physical Attack Detection in Water Distribution Systems

Water Distribution Systems are essential infrastructures that ensure reliable water supply. The integration of cyber components introduces vulnerabilities to Cyber-Physical Attacks. This paper presents the Rectilinear Hybrid Belief Classifier Network, a hybrid machine learning framework designed to detect and reduce Cyber-Physical Attacks in real-time. The Rectilinear Hybrid Belief Classifier Network utilizes Mean-Standard Deviation scaling and a unique Rectilinear Transformation Technique to preprocess and reduce data dimensionality effectively. This enhances the system’s resilience against noisy, high-dimensional data. Autoencoders for feature compression and Deep Belief Networks for complex pattern recognition are combined in the Deep Hybrid-Encoder Belief Network for anomaly detection. This approach enables the Rectilinear Hybrid Belief Classifier to identify cyber-physical anomalies with 99% accuracy and 96% precision. Comparative results shows that the Rectilinear Hybrid Belief Classifier framework excels the existing models in accuracy, scalability, and adaptability to operational noise, making it an ideal solution for secure, large-scale Water Distribution Systems operations.