UDC 004.93; UDC 004.896
KINEMATIC MODEL FOR GENERATING REFERENCE DYNAMIC PARAMETERS OF A USER’S SIGNATURE
A. H. Tancerov, Postgraduate Student, Department of «Programming», Penza State Technical University (PenzGTU), Penza, Russia;
orcid.org/0009-0006-7695-0514, e-mail: This email address is being protected from spambots. You need JavaScript enabled to view it.
E. A. Danilov, PhD. in Engineering, Associate Professor of the Department of «Programming», Penza State Technical University (PenzGTU), Penza, Russia;
orcid.org/0000-0003-4114-7036, e-mail: This email address is being protected from spambots. You need JavaScript enabled to view it.
One of the approaches to enhance the reliability of biometric user identification, based on the standard ization of dynamic signature implementations, is considered. Identification and authentication as key pro cesses in information security ensure the verification of user's identity and their right to access system, data, or resources. Biometric identification is carried out through registration, recording, and mathematical en coding of biometric samples (both constant physiological traits and variable behavioral characteristics), which allows the provided sample to be compared—using «one-to-many» principle—with data stored in the system, while verification follows «one-to-one» principle. Particular attention is paid to the issue of variabil ity of behavioral traits and low discriminative capability of static signatures, which complicates their use for authentication. The aim of this work is to develop and analyze a method for the standardization of dynamic signatures based on a kinematic model that assumes approximation of additional dynamic parameter values dependence on the tuple of standard parameters. This approach allows, by knowing the characteristics of analytically constructed dependency, the formation of a partially synthetic full sample from each reduced sample, which is then used for training user identification models. The model is implemented as two- or three-layer neural network with a large number of neurons in hidden layers, and it provides for the cluster ing of several stable yet different signature realizations using a SOM network. The study confirms the effi ciency of the proposed method, its relevance for practical applications, and the prospects for further improv ing the procedures for storing and utilizing dynamic signature curve samples under real-time conditions.
Key words: : dynamic signature, identification, user authentication, neural networks, Kohonen map, standardization, biometric authentication, neural network algorithms.