Software Development for Detection of non-Standard User Behavior

The ways of improving the cybersecurity level of information systems are investigated in the diploma. Analysis of the internal users’ actions is proposed as a possible solution, because insiders’ threats become more powerful and harmful. The development of a system for identifying suspicious users’ actions continues to be an urgent issue. In order to retrieve methods, that could effectively solve the problem of identifying non-standard user behavior, various known anomaly-detection approaches based on machine learning algorithms were explored in the current study. Some of the approaches, which showed good results during the tests, were combined in the proposed method of detection of non-standard user behavior. All the analysis and tests were performed on the dataset, which represent the emails of the real users. The volume of the diploma is 56 thousand characters, includes 14 illustrations and 50 references.