V.A. Ivanov, A.A. Smirnov, D.A. Nikolaev

Statement of the problem. A significant place in biometric identification systems is given to face recognition based on twodimensional images using artificial neural networks. One of the assessments of the functional effectiveness of such systems, along with the detection properties, is the resistance to destabilization, which is expressed in errors of the 1st and 2nd kind.

Goal. To investigate artificial neural networks for stability to destabilization by statistical testing methods.

Results. A method for estimating errors of the 2st kind for artificial neural networks under the influence of destabilizing factors - the presentation of identifiers of different people - has been developed and tested.

Practical significance. Using the example of two artificial neural networks, the values of errors of the 2st kind are obtained. The laws of distribution of such errors are selected.

Ivanov V.A., Smirnov A.A., Nikolaev D.A. Noise immunity of artificial neural networks in the recognition of images of people's faces. Radiotekhnika. 2022. V. 86. № 1. P. 61−65. DOI: https://doi.org/10.18127/j00338486-202201-10 (In Russian)

1. Rukovodstvo po sostavleniju specifikacij na SKUD. M.: Security Focus. 2017. 170 s. (In Russian).
2. Magauenov R. Sistemy ohrannoj signalizacii: Osnovy teorii i principy postroenija: Ucheb. posobie dlja vuzov. M.: Gorjachaja linija − Telekom. 2017. 494 s. (In Russian).
3. Gmurman V.E. Teorija verojatnostej i matematicheskaja statistika: Ucheb. posobie dlja vuzov. M.: Vysshaja shkola. 2003. 479 s. (In Russian).
4. Dvojris L., Ivanov V., Krjukov I. Realizacija kriteriev obnaruzhenija i razlichenija signalov v srede MathCAD. Radiotehnika. 2021. № 2 (In Russian).