Polhul T. Information technology for fraud detection during mobile applications installation using data mining

This dissertation research is dedicated to developing the informational technology to detect fraud during mobile application installation using data mining. The purpose of the dissertation research is to increase the accuracy and speed of fraud detection processes during mobile application installation. For the first time, a method of overcoming the heterogeneity of input data is proposed, which is a set of procedures for feature selecting, dimensionality reduction and data normalization, the difference of which lies in the new model of overcoming the heterogeneity of data by scaling information, which allows the whole set of heterogeneous user data to be reduced to a vector, reducing the diagnostic value of information. The model of users' classification based on deep neural networks in terms of dimensionality reduction and data normalization has been improved according to the proposed method of overcoming heterogeneity of data, which is the basis for creating general fraudsters fingerprint in order to simplify their detection processes. For the first time, a generalized method for fraud detection in during installation of mobile applications has been developed, the difference being the use of the proposed users' classification model and the method of overcoming the heterogeneity of the input data, which allows defining users’ classes and increasing the reliability of fraud detection during mobile app installs. The results of the qualification research paper were implemented at foreign (Garuda AI B.V.) and Ukrainian (LLC «WinInteractive», LLC «4HighTech», PE «Litsoft») enterprises and to the educational process (at the Computer Science Department of Vinnytsia National Technical University and at the Department of Informatics, Software Engineering and Economic Cybernetics of Kherson State University). The implementation of the results of dissertation research is confirmed by the relevant acts.