Article 5321

Title of the article

Secure-design artificial intelligence: the advantages of replacing classical artificial neurons with neurons of geometric mean and harmonic mean 

Authors

Vitaliy S. Lukin, Junior researcher, Regional Training and Research Center of “Information security”, Penza State University (40 Krasnaya street, Penza, Russia), E-mail: ibst@pnzgu.ru
Aleksandr I. Ivanov, Doctor of engineering sciences, associate professor, scientific adviser, Penza Scientific Research Electrotechnical Institute (9 Sovetskaya street, Penza, Russia), E-mail: ivan@pniei.penza.ru
Konstantin A. Perfilov, Principal specialist of the Center of information technologies in communication and data protection, Administration of the Ministry of Internal Affairs of the Russian Federation in Penza region (159 Pushkina street, Penza, Russia), E-mail: perfilov58@gmail.com 

Index UDK

004.8, 004.032.26 

DOI

10.21685/2072-3059-2021-3-5 

Abstract

Background. Responsible artificial intelligence applications must run in a secure execution. One of the directions for creating such applications is the use of large artificial neural networks, for example, consisting of perceptrons automatically trained in accordance with State Standart R 52633.5. The purpose of the work is to replace classical perceptrons with artificial neurons of the geometric mean and harmonic mean. Materials and methods. Perceptrons enrich biometric data through their accumulation in linear space, that is, they can be considered as artificial neurons of the arithmetic mean. Artificial chi-square neurons and Mahalonobis neurons accumulate data in the space of the standard deviation. The transition to other classes of artificial neurons by formal replacement of the data accumulation space with the geometric mean and harmonic mean is considered. Results. The advantage of the geometric mean and harmonic mean neurons is that their software implementation does not require preliminary calculation of the first statistical moments of small samples of biometric data. The latter removes the problem of hiding the statistical moments of the “Friend” image through encryption of the tables’ connections and weight coefficients of the artificial neurons network. Conclusions. The use of artificial neurons of the geometric mean and harmonic mean is a promising direction for creating neural network decision rules of artificial intelligence in a performance protected from knowledge extraction. 

Key words

artificial neurons, harmonic mean neurons, geometric mean neurons, artificial intelligence protection 

Download PDF
References

1. Ivanov A.I., Zakharov O.S. Sreda modelirovaniya «BioNeyroAvtograf» [Programmnyy produkt sozdan laboratoriey biometricheskikh i neyrosetevykh tekhnologiy, razmeshchen s 2009 g. na sayte AO «PNIEI»] = Modeling environment "BioNeuroAutograph" [The software product was created by the laboratory of biometric and neural network technologies, posted in 2009 on the website of JSC “PNEI”]. (In Russ.). Available at: http://pniei.rf/activity/science/noc/bioneuroautograph.zip
2. Ivanov A.I. Avtomaticheskoe obuchenie bol'shikh iskusstvennykh neyronnykh setey v biometricheskikh prilozheniyakh: ucheb. posobie = Automatic training of large artificial neural networks in biometric applications: textbook. Penza, 2013:30. (In Russ.). Available at: http://pniei.rf/activity/science/noc/tm_IvanovAI.pdf
3. Yazov Yu.K., Volchikhin V.I. [et al.]. Neyrosetevaya zashchita personal'nykh biometricheskikh dannykh = Neural network protection of personal biometric data. Moscow: Radiotekhnika, 2012:157. (In Russ.)
4. Ivanov A.I., Kupriyanov E.N. Zashchita iskusstvennogo intellekta: ortogonalizatsiya statistiko-neyrosetevogo analiza malykh vyborok biometricheskikh dannykh: preprint = Artificial intelligence protection: orthogonalization of statistics-neural network analysis of small samples of biometric data: preprint. Penza: Izd-vo PGU, 2020:72. (In Russ.)
5. Volchikhin V.I., Ivanov A.I. Neural network molecule: solving the inverse problem of biometrics through software support of quantum superposition at the outputs of a network of artificial neurons. Vestnik Mordovskogo universiteta = Bulletin of Mordovian University. 2017;27(4):518–523. (In Russ.)
6. Tekhnicheskaya spetsifikatsiya (proekt, publichnoe obsuzhdenie chlenami TK 26 zakanchivaetsya v 2020 godu) «Kriptograficheskaya zashchita informatsii. Zashchita neyrosetevykh biometricheskikh konteynerov s ispol'zovaniem kriptograficheskikh algoritmov » = Technical specification (draft, public discussion by TC 26 members ends in 2020) “Cryptographic information protection. Protection of neural network biometric containers using cryptographic algorithms”. Penza, 2020. (In Russ.)
7. Dodis Y., Reyzin L., Smith A. Fuzzy Extractors: How to Generate Strong Keys from Biometrics and Other Noisy. Proc. EUROCRYPT. 2004;April 13:523–540.
8. Monrose F., Reiter M., Li Q., Wetzel S. Cryptographic key generation from voice. Proc. IEEE Symp. on Security and Privacy. 2001:202–213.
9. Ramírez-Ruiz J., Pfeiffer C., Nolazco-Flores J. Cryptographic Keys Generation Using FingerCodes. Advances in Artificial Intelligence. IBERAMIA-SBIA, 2006:178–187.
10. Volchikhin V.I., Ivanov A.I., Malygina E.A., Yunin A.P. The ratio of neurons’ power with linear and quadratic biometric data enrichers. Izvestiya vysshikh uchebnykh zavedeniy. Povolzhskiy region. Tekhnicheskie nauki = University proceedings. Volga region. Engineering sciences. 2018;(1):17–25. (In Russ.)
11. Malygina E.A. Biometriko-neyrosetevaya autentifikatsiya: perspektivy primeneniya setey kvadratichnykh neyronov s mnogourovnevym kvantovaniem biometricheskikh dannykh: preprint = Biometrical neural network authentication: prospects for the application of quadratic neuron networks with multilevel quantization of biometric data: preprint. Penza: Izd-vo PGU, 2020:114. (In Russ.)
12. Ivanov A.I., Bezyaev A.V., Malygina E.A., Serikova Yu.I. The second national standard of Russia for fast automatic training of large artificial neural networks on small samples of biometric data. Bezopasnost' informatsionnykh tekhnologiy: sb. nauch. st. po materialam I Vseros. nauch.-tekhn. konf. = Information technology security: proceedings of the 1st All-Russian scientific and engineering conference. Penza, 2019:174–177. (In Russ.)
13. Perfilov K.A. A geometric mean criterion used to test the reliability of statistical hypotheses of the distribution of biometric data. Bezopasnost' informatsionnykh tekhnologiy: tr. nauch.-tekhn. konf. klastera penzenskikh predpriyatiy = Information Technology Security: proceedings of scientific and technical conference cluster of Penza enterprises. Penza, 2014;9:92–93. (In Russ.)
14. Akhmetov B.S., Ivanov A.I., Perfilov K.A. Using the geometric mean, expected and observed probability functions as a statistical criterion for assessing the quality of biometric data. Trudy Mezhdunarodnogo simpoziuma Nadezhnost' i kachestvo = Proceedings of the International symposium Reliability and Quality. Penza, 2015;2:281–283. (In Russ.)
15. Perfilov K.A., Ivanov A.I., Protsenko E.D. Expanding the variety of statistical criteria used to test the hypotheses of the biometric data values distribution. Evropeyskiy soyuz uchenykh = European union of scientists. 2015;(4–5):9–12. (In Russ.)
16. Ivanov A.I., Perfilov K.A., Malygina E.A. Multivariate statistical analysis of the quality of biometric data on extremely small samples using the geometric mean criteria calculated for the analyzed probability functions. Izmerenie. Monitoring. Upravlenie. Kontrol' = Measurement. Monitoring. Administration. Control. 2016;(2):58–66. (In Russ.)
17. Ivanov A.I., Bannykh A.G., Kupriyanov E.N. A collection of artificial neurons equivalent to statistical criteria for their combined use in testing the hypothesis of normality of small samples of biometric data. Bezopasnost' informatsionnykh tekhnologiy: sb. nauch. st. po materialam I Vseros. nauch.-tekhn. konf. = Information technology security: proceedings of the 1st All-Russian scientific and engineering conference. Penza, 2019:163–172. (In Russ.)
18. Ivanov A.I., Perfilov K.A, Lukin V.S. Neural network generalization of a family of statistical criteria for geometric mean and harmonic mean for precision analysis of small samples of biometric data. Informatsionno-upravlyayushchie telekommunikatsionnye sistemy, sredstva porazheniya i ikh tekhnicheskoe obespechenie: sb. nauch. st. Vseros. nauch.-tekhn. konf. = Information and administration telecommunication systems, weapons and their technical support: proceedings of the All-Russian scientific and engineering conference. Penza: NPP «Rubin», 2019:50–63. (In Russ.)
19. Marshalko G.B. On the security of a neural network-based biometric authentication scheme. Matematicheskie voprosy kriptografii = Mathematical questions of cryptography. 2014;5:87–98.
20. Bogdanov D.S., Mironkin V.O. Data recovering for a neural network-based biometric authentication scheme. CTCrypt. Preproceedings. 2018:262–273. Available at: https:// ctcrypt.ru/files/2018/23_Bogdanov.pdf
21. Bezyaev A.V. Biometriko-neyrosetevaya autentifikatsiya: obnaruzhenie i ispravlenie oshibok v dlinnykh kodakh bez nakladnykh raskhodov na izbytochnost': preprint = Biometrical neural network authentication: detecting and correcting errors in long codes without the overhead of redundancy: preprint. Penza: Izd-vo PGU, 2020:40. (In Russ.)

 

Дата создания: 09.12.2021 08:47
Дата обновления: 09.12.2021 09:35