Artificial Intelligence Technologies as a Tool for Ensuring the Cybersecurity of Critical Information Infrastructure

The application of artificial intelligence and machine learning technologies for ensuring the cybersecurity of critical information infrastructure objects has been researched. A methodology for training machine learning models for the detection and classification of cyber incidents has been developed and generalized. Approaches to the formulation of a family of models and methods for representing events in a multidimensional feature space have been considered. The use of linear decision boundaries and logistic (sigmoid) functions for estimating the prob-ability of fraudulent events is described. Loss functions for classification and regression tasks are analyzed, in particular, the sum of squared errors and negative logarithmic likelihood. First- and second-order optimization algorithms are researched, including gradient descent, stochastic gradient descent, and its variations of AdaGrad and Adam. The principle of the gra-dient descent method is demonstrated, and practical recommendations are given for selecting optimization algorithms depending on the type of task and data size. An example of the pro-posed methodology application for analyzing transactions in the energy sector to detect fraud is given.