Estimation of failure on-demand probability and malfunction rate values in cyber-physical systems of nuclear power plants

Nuclear Power Plants (NPPs) are making increasing use of digital Instrumentation and Control (I&C) systems, which makes them Cyber-Physical Systems (CPSs). In CPSs, cyber and physical processes are dependent and interact with each other: sensors, actuators, communication and computational units are all interconnected to realize functionalities of real-time monitoring, dynamic control and decision support, for normal operation as well as in case of accidents. However, an emerging concern is that the use of computer-based technologies might increase the exposure to failures and accidents, providing new channels for their initiation and propagation. System integrity can be, indeed, affected by hardware component failures, human errors, communication malfunctions and software errors, but also compromised by security breaches and cyber attacks. In practice, these latter could be confused with random components failures on-demand and malfunctions, misjudging their actual nature of malicious cyber attacks and, thus, leading to wrong counteractions. In this study, we analyze and model stochastic failures in components of CPSs, with the purpose of estimating reference values of failure on-demand probabilities and malfunction rates. Considering these as true values, then, significant difference with statistical estimates from field data collected on the real CPS can be used to detect malicious attempts at altering the safety of a NPP. A digital I&C system of a NPP is taken as illustrative case study, in which components stochastic failures resulting in different system responses are analyzed, and Fault Tree Analysis (FTA) and Markov Chain Modeling (MCM) are taken as approaches to estimate the reference failure on-demand probabilities and malfunction rates.