Open AccessMODEL OF FORMATION OF STUDY EXAMPLES OF THE NEURAL NETWORK INTENDED FOR THE ANALYSIS OF THE KEYBOARD HANDWRITING
Author(s) -
Liudmyla Tereikovska
Publication year - 2020
Publication title -
kìberbezpeka. osvìta, nauka, tehnìka
Language(s) - English
Resource type - Journals
ISSN - 2663-4023
DOI - 10.28925/2663-4023.2020.9.104114
Subject(s) - computer science , handwriting , convolutional neural network , field (mathematics) , artificial neural network , intelligent character recognition , password , keystroke logging , handwriting recognition , reading (process) , biometrics , covert , state (computer science) , set (abstract data type) , artificial intelligence , pattern recognition (psychology) , human–computer interaction , speech recognition , feature extraction , computer security , algorithm , linguistics , philosophy , mathematics , character recognition , political science , pure mathematics , law , image (mathematics) , programming language
The article is devoted to increasing the efficiency of technologies of covert monitoring of operators' activity by information and control systems of various purposes for face recognition and emotional state. It is shown that from the standpoint of the possibility of using standard computer peripherals as a sensor for reading biometric parameters, inalienability from the user, the widespread use of information control systems of symbolic password and technological data, the complexity of forgery of biometric information, and the possibility of covert monitoring prospects have the means of keyboard analysis. The necessity of improving the methodology of neural network analysis of keyboard handwriting for authentication and recognition of the emotional state of information computer system operators is substantiated. The prospects of application of convolutional neural networks are determined, which leads to the need to improve the technology of determining the parameters of educational examples in terms of forming the input field of convolutional neural network and forming many parameters of keyboard handwriting to be analyzed. A model of formation of educational examples has been developed, which due to the application of a reasonable set of input parameters and the use of a rectangular input field of a convolutional neural network reduces the resource consumption of neural network recognition tools and provides accuracy of neural network analysis of keyboard handwriting at 75%. The proposed theoretical solutions were verified by computer experiments. The expediency of correlation of ways of further researches with development of representative databases of keyboard handwriting is shown.