USER IDENTIFICATION METHOD BY MEANS OF BIOMETRIC IMAGE OF KEYSTROKE DYNAMICS WITH DOUBLE-CHAINED REPRESENTATION

Abstract

To date, more and more relevant is the task of establishing the authorship of an electronic document or typed on the keyboard text. Among the many options for solving this problem, a method based on biometric data obtained as a result of the analysis of the keystroke dynamics is highlighted. In particular, the paper considers the main parameters obtained on the basis of time stamps of keystrokes, as well as the method from the formalized doubly-chained representation. A method for identifying biometric images based on a statistical algorithm that forms a certain type of trust area within which the parameters of the identified object will be located has been described and implemented. In addition, the paper describes methods for preliminary processing of data (emission filtering, normalization and clustering), as well as their effectiveness based on the evaluation of the accuracy of crossvalidation for different values of the access threshold. Thus, in the course of practical research, a simple probabilistic algorithm was implemented to identify samples of the keyboard handwriting with a pass phrase, an estimation of the identification accuracy was made, the necessity of preliminary processing of input data for improving the quality of identification was justified. Based on the results obtained, it is further assumed that a complete biometric image of the keyboard handwriting will be formed, with further identification by arbitrary text.