Model of the "biometry-code" converter based on artificial neural networks for analysis of facial thermograms
Модель преобразователя «биометрия-код» на основе искусственных нейронных сетей для анализа термограмм лица субъектов

Pavel Lozhnikov, Omsk State Technical University, Russian Federation, Samal Zhumazhanova, Omsk State Technical University, Russian Federation
2021 Digital Technology Security  
Existing asymmetric encryption algorithms involve the storage of a secret private key, authorized access to which, as a rule, is carried out upon presentation of a password. Passwords are vulnerable to social engineering and human factors. Combining biometric security techniques with cryptography is seen as a possible solution to this problem, but any biometric cryptosystem should be able to overcome the small differences that exist between two different implementations of the same biometric
more » ... ameter. This is especially true for dynamic biometrics, when differences can be caused by a change in the psychophysiological state of the subject. The solution to the problems is the use of a system based on the "biometrics-code" converter, which is configured to issue a user key after presentation of his/her biometric image. In this case, the key is generated in advance in accordance with accepted standards without the use of biometric images. The work presents results on using thermal images of a user for reliable biometric authentication based on a neural network "biometrics-code" converter. Thermal images have recently been used as a new approach in biometric identification systems and are a special type of biometric images that allow us to solve the problem of both the authentication of the subject and the identification of his psychophysiological state. The advantages of thermal imaging are that this technology is now becoming available and mobile, allowing the user to be identified and authenticated in a non-contact and continuous manner. In this paper, an experiment was conducted to verify the images of thermograms of 84 subjects and the following indicators of erroneous decisions were obtained: EER = 0.85 % for users in the "normal"state.
doi:10.17212/2782-2230-2021-2-154-165 fatcat:lyrtjv76rrejlhqnutxxlxli2u