МЕТОД ОБЕСПЕЧЕНИЯ БЕЗОПАСНОСТИ ПЕРСОНАЛЬНЫХ ДАННЫХ ПРИ ИХ ОБРАБОТКЕ В ИНФОРМАЦИОННОЙ СИСТЕМЕ НА ОСНОВЕ АНАЛИЗА ПОВЕДЕНИЯ ПОЛЬЗОВАТЕЛЕЙ

Abstract

Introduction: As information technologies develop, new classes of personal data protection software come out. One of them is User Behavior Analytics. In the development of such systems, machine learning methods are widely used, including the applications of artificial neural network theory. However, the approaches to protection software development based on machine learning have not been studied well enough. Purpose: Developing a method of creating an artificial neural network which would provide the analysis of authorized behavior of information system users and the detection of abnormalities in their behavior signaling about criminal activity. Results: A review of the approaches to provide information security with artificial neural networks demonstrates their active development in several directions, including the detection of abnormalities. A new method to create an artificial neural network has been developed, including the proposals about determining the network type, the range of numeric values for the input and output signals, the number of the layers and neurons in a layer, the learning method, and the type of the activation functions. As the input values, user behavior characteristics can be used, namely: a set of the user's data, the point of access to the information system, the set of user's actions, the time of the access or time of certain actions, the general duration of runtime operations. With user's access time as an example, an approach has been proposed to assign numeric values to the characteristic of a user, based on fuzzy set theory application. Practical relevance: A trained neural network provides a more efficient detection of abnormalities in user behavior than an information security specialist without special automation tools.