Data Distribution-Based Change Detection Framework in SWaT Security Monitoring

Networks of sensors and Internet of Things have led to an ever-increasing amount of data that is now more commonly available in streaming settings. Often it is assumed that the process generating these streams is stationary, however, in real world scenarios, systems are evolving and dynamic in nature. This results in degradation of trained model predictions and decision making process. Thus, methods and approaches to be able to detect when there is a change or drift in the environment are necessary. In this research project, we proposed a change detection framework and analyzed various data distribution-based change detection algorithms for real-time analysis and change detection in Secure Water Treatment. Results show that the framework is promising, able to effectively monitor the cyber-physical system, and out of the algorithms we experimented with, the Kolmogorov-Smirnov WINdowing algorithm performed better in terms of changes detected and detection delay, albeit with a high false alarm rate.