INFORMATION TECHNOLOGY FOR THE CONSTRUCTION OF INTELLIGENT SYSTEMS FOR DETECTING FRAUD DURING MOBILE APPLICATIONS INSTALLATION

Abstract

Information technology for the construction of intelligent systems for detecting fraud during mobile applications installations, which is desirable to use in developing such a class of systems, is proposed in this paper. The intelligent processing of available data by the user is done. The scaling which is based not on the value of the feature, but on the end-point information of the feature, is proposed based on this intelligent processing of the data. A system with an intellectual component - the formation of a knowledge base that will allow fraudsters to be identified and which will include anomaly analysis rules - was proposed, so that the emergence of a new anomaly in the data allows for the creation of a new rule. Such knowledge base can be expanded due to the possibility of an emergence of a new kind of anomaly in data (fraud). The received set of rules will allow creating a generalized fraudster's fingerprint, noting even the new and unknown for experts fraudulent patterns, based on the algorithms developed in the work. The classification of suspicious users to a class of fraudsters or organic users is made using fuzzy logic. The information technology for the construction of intelligent systems that will be able to adapt to the emergence of new types of fraud is proposed based on the proposed intelligent processing of available user data, the scaling by end-point information, and the development of knowledge bases. According to the tasks which should be solved by such intelligent systems, their structure is proposed: subsystem of user data characteristics identifying; subsystem of overcoming heterogeneity; subsystem of classification model training; subsystem of classification; subsystem of fraudsters database formation; subsystem of knowledge base (for detecting fraud) formation; subsystem of data mining and user patterns formation; subsystem of general fraudster fingerprint prediction. The proposed information technology for the construction of intelligent systems allows processing of various input data, which in the process gives the opportunity to form a generalized fraudster fingerprint.