Specific features of operational assessment of security of critical resources based on adaptive neural network filtering

The object of research is a new methodological approach to adaptive neural network filtering as a mathematical tool for improving the accuracy and efficiency of evaluating some properties of complex technical systems. This approach is one of the options for the practical application of adaptive (hybrid) filtering methods. The analysis of the features of this approach determining the rationality of its application for the operational assessment of the security of critical resources is carried out. The theoretical aspects of the application of a hybrid adaptive approach to the operational assessment of the security of critical resources, combining traditional methods of Kalman filtering with the capabilities of artificial neural networks with training, are considered. The analysis of the features of this approach is carried out, which allows learning and adjusting the weighting coefficients of filtering to the statistical characteristics of the indicators of the security of critical resources, measured and observed both linearly and non-linearly. A sequence of calculations and analytical expressions are proposed for calculating the estimated values of auxiliary indicators of the state of security indicators based on an adaptive hybrid filter containing a trainable artificial neural network. The approach assumes the practical possibility of operational assessment of the security of critical resources using adaptive hybrid filtering of random processes that characterize the dynamics of changes in the state variables (indicators) of the security of such resources at a certain time interval. It takes into account the uncertainty of the initial data, incompleteness and vagueness of a priori information about the statistics of security indicators and surveillance noise. At the same time, the proposed approach makes it possible to obtain estimates adequate to the tasks of operational security control and, ultimately, works out to increase the reliability of information security control of modern critical resources.