Choice of indicators for forecasting cyber protection of computer systems

Abstract

The article proposes an algorithm for selecting indicators for predicting cyber security of computer systems. Cyber defense processes are random multidimensional, dynamic non-stationary, active (targeted), which complicates the task of forecasting cyber security indicators. The analysis of publications showed the complexity of choosing the most effective method of forecasting cyber security, which consists in determining the classification of methods of forecasting the characteristics of each method, the list of requirements for retrospective information. Thus, the use of extrapolation in forecasting always involves the use of any models, so modeling is the basis for extrapolation. Forecasting is a rather complex task, which is confirmed by the analysis of the causes and factors that potentially affect changes in the forecasted indicator. Solving such a task, like any other complex task, requires a systematic approach that helps to understand the essence of the problem and choose adequate methods of solving it, as well as assess the reasons for possible failures. The resulting algorithm contains a multi-story model, both in the class of linear and in the class of nonlinear models according to input variables; exclusion of individual members of a better private description and based on this expansion of the basic set of arguments; is optimal in terms of computational costs for the iterative algorithms of the method of the group computational algorithm of the sliding test criterion calculation scheme. And it also has the ability to estimate coefficients in models both by the method of least squares and by the method of least modules.