Identifying anomalous system behavior is the key to detect any cyber-physical attacks on any water distribution systems. Analyzing system data and understanding the system and the relationships between different components of the system can provide a basis to develop a hybrid method to identify these anomalous behaviors. Lack of in-depth knowledge in system hydraulics and control may fail to identify the anomalous behavior with accuracy. Hence, the implementation of system hydraulics and control in attack detection algorithm is important. This paper presents an algorithm that not only considers the concept of data mining but also the hydraulics and system relationships. Three specific modules namely statistics and pattern recognition module, principal component analysis module, and hydraulics and system relationships module are included in the algorithms to detect any cyber-physical attack on a literature network known as C-Town. The evaluation criteria have been calculated for three of these modules separately. Results show that the hydraulics and system relationship module can perform better than any other modules considered in the study. The performance score for the attack detection and the score for the confusion matrix for this module are respectively 1.0 and 0.86 resulting a combined score of 0.93.
展开▼
