Data-driven failure analysis for the cyber physical infrastructures

Digital transformation is a main driver of a modern approach to providing cyber safety and ecurity in the environment of smart systems: smart home, IIoT, smart building, smart megapolis, VANET, FANET, WSN, etc. For traditional computer networks, security was to ensure confidentiality, integrity, and availability of data. The last decade, with the advent of dynamic machine-to-machine infrastructures, the goal is forced to be ensured in safety and reliability of physical processes at cyber system. Due to the increased mobility of topology and the growing amount of data (Big Data) undergoing the processing, traditional methods of system analysis become ineffective, so the researchers are faced with the task of creating new methods for ensuring cyber security that meet new challenges. This paper outlines the essence of the approach to the detection of weaknesses caused by failures of the components of cyber physical infrastructure. The paper discusses a method of data-driven analysis for failure detection and prediction. The proposed technique is based on the modified 'k' method of nearest neighbors (KNN) extended with application of Dempster-Shafer (DS) theory and our suggestion to estimate the spatial-temporal correlation of the connected devices in the cyber physical environment. Our method shows above 99%-level of effectiveness comparing to common fault management approaches.