Metadata (cookie, IP address, location) may be collected for the functioning of the site. If you do not want this data processed, then you should leave the site.
OK
AICyone, LLC
AICyone, LLC

We develop artificial intelligence systems and offer software for their rapid development and implementation. Our team also conducts scientific research and data analysis work at the Omsk State Technical University.

Move to the team description.



Team
Our team consists of young and experienced software developers, specialists, and scientists in the fields of Data Science, ML, and AI.
  • Alexey Sulavko
    CEO
    scientific director of projects
    Doctor of Technical Sciences, Professor of the complex information security department at Omsk State Technical University. Expert from Russia in the International Technical Committee ISO/IEC JTC 1/SC 42 "Artificial intelligence". Reviewer in international scientific journals (published by Elsevier, Nature Publishing Group, MDPI, etc.), author of more than 150 scientific publications in the field of AI.
    ORCID: 0000-0002-9029-8028
    Scopus Author ID: 56825944600
    ResearcherID: N-3388-2016
    SPIN-код (e-library): 7455-0834
    IEEE Profile
    Omsk State Technical University profile
  • Denis Stadnikov
    data scientist,
    software architect
    Postgraduate student at Omsk State Technical University,
    author of more than 30 scientific publications in the field of AI. More than 4 years of experience in data science. Research interests: machine learning, artificial intelligence, pattern recognition, artificial neural networks, neural interfaces
  • Adil Choban
    data scientist,
    DevOps
    Postgraduate student at Omsk State Technical University,
    author of more than 20 scientific publications in the field of AI. More than 4 years of experience in data science. Research interests: machine learning, pattern recognition, artificial intelligence, artificial neural networks, neural interfaces
  • Alexander Samotuga
    AI/ML developer,
    translator
    Candidate of Technical Sciences, Associate Professor of the department. comprehensive information protection of Omsk State Technical University. Professional translator in the field of information security. The author of more than 35 scientific publications in the field of AI.
    ORCID: 0000-0001-5647-7498
    Scopus Author ID: 56568558300
    SPIN-код (e-library): 4235-1670
    Omsk State Technical University profile
  • Pavel Lozhnikov
    PR manager,
    scientific consultant
    Doctor of Technical Sciences,
    Head of department comprehensive information protection Omsk State Technical University
    Certified PMVoK specialist. Author of more than 150 scientific publications.
    ORCID: 0000-0001-7878-1976
    Scopus Author ID: 55027255900
    ResearcherID: I-3666-2013
    SPIN-код (e-library): 4324-2543
    Omsk State Technical University profile
  • Irina Panfilova
    AI/ML developer,
    conceptual designer

    Postgraduate student at SamSTU,
    author of 10 scientific publications in the field of AI. More than 4 years of experience in data science.
    Research interests: machine learning, biometrics, artificial intelligence, pattern recognition, artificial neural networks, biometric-to-code converters



  • Georgy Suvyrin
    Backend-developer
    Master's student at Omsk State Technical University,
    More than 4 years of experience in industrial and commercial development. Scientific interests: analysis of telemetric data for predictive diagnostics of equipment
  • Daniil Inivatov
    AI/ML developer
    Postgraduate student at Omsk State Technical University,
    He is the winner and prize-winner of more than 20 international mathematical olympiads. Author of more than 40 scientific publications. Scientific interests: machine learning, artificial intelligence, biometric-code converters, assessment of psychophysiological state
  • Anastasia Serikova
    project manager,
    creative designer
    Creative UI Developer
  • Yuri Varkentin
    Frontend-developer

    Web Application Developer
  • Yuri Dorogov
    Frontend-developer

    Web Application Developer
  • Nikita Shagov
    full stack developer

    Web Application Developer

Omsk Scientific School
and history of the AICyone LLC enterprise
The company AI ZION LLC (AICyone, LLC) was founded in 2021 in the city of Omsk by a group of specialists and scientists in the field of machine learning and artificial intelligence (AI). We are followers of the scientific school of Boris Nikolaevich Epifantsev (1939–2016), Doctor of Technical Sciences, professor, leading scientist of Omsk. Epifantsev B.N. developed many scientific areas (remote biometric identification, detection of unauthorized access to product pipelines based on vibro-acoustic monitoring and visual control, information leakage protection systems, etc.), a feature of which is the use of Bayesian probabilistic inference networks to solve problems of data mining and non-destructive control. The ideas of Epifantsev B.N. served as the foundation for the creation of the Omsk scientific school of machine learning and were developed in the works of his students. At the moment, the scientific school is based at one of the leading universities in Russia, Omsk State Technical University (OMSTU), at the Department of Integrated Information Security (KIP), where active scientific work is carried out in the following main areas:
  • Machine learning using small data samples
  • Ensembles of machine learning models
  • Pattern recognition, taking into account the correlation properties of the feature space
  • Highly reliable biometric authentication and neural interfaces based on electroencephalography Artificial neural networks
  • Bio-inspired machine learning models and algorithms (artificial immune systems)
  • Protected execution mode of neural network models and AI algorithms

We realized that we needed a tool for scientific research that could combine all of these areas. This is how the AIConstructor project was born, within which we created a system for quickly testing hypotheses in the field of machine learning. Then we realized that this project represented something more than just a software package for research and created our own IT company.

Publications of the research team
Our team has prepared more than 200 publications on the ML, AI topics.
We also have patents.
Personal Identification Based on Acoustic Characteristics of the Outer Ear Using Cepstral Analysis, Bayesian Classifier and Artificial Neural Networks
cepstrograms, windowed Fourier transform, Bayes formula, multilayer neural networks, acoustic signal, pattern identification, machine learning, echograms, biometrics, information security, ear structure
https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/bme2.12037
Sulavko, A. E., Samotuga A.E., Kuprik I.A. Personal Identification Based on Acoustic Characteristics of the Outer Ear Using Cepstral Analysis, Bayesian Classifier and Artificial Neural Networks // IET Biometrics. - 2021. - p. 1-14 (early view)

A hypothesis is discussed concerning the use of echograms of the external auditory canal for personal identification. The authors have developed a device for measuring the acoustic parameters of the external auditory canal. Obtained echograms can be used as biometric patterns for identification and authentication of subjects. Two types of biometric parameters are considered based on spectral and cepstral analyses of echograms. The authors used two approaches for recognizing ear patterns: the first was based on Bayes' formula and the second on artificial neural networks (convolutional and fully connected). The Bayesian classifier has been found to show a lower percentage of identification errors with an equal error rate (EER) = 0.0053. The best result for neural networks was EER = 0.0266. An experiment the authors repeated with the same subjects six months after the initial data collection showed insignificant deviation in the number of wrong decisions (EER = 0.008).
An abstract model of an artificial immune network based on a classifier committee for biometric pattern recognition by the example of keystroke dynamics
biometric authentication, bagging, boosting, feature subspaces, machine learning on small samples, ensembles of models.
http://www.mathnet.ru/links/e12035696decffaaf458449f2cfe2a48/co853.pdf
Sulavko AE. An abstract model of artificial immune network based on classifiers committee for biometric patterns recognition on the example of keystroke dynamics. Computer Optics 2020. – Т. 44, № 5. – С. 830-842. – DOI: 10.18287/2412-6179-CO-717

An abstract model of an artificial immune network (AIS) based on a classifier committee and robust learning algorithms (with and without a teacher) for classification problems, which are characterized by small volumes and low representativeness of training samples, are proposed. Evaluation of the effectiveness of the model and algorithms is carried out by the example of the authentication task using keyboard handwriting using 3 databases of biometric metrics. The AIS developed possesses emergence, memory, double plasticity, and stability of learning. Experiments have shown that AIS gives a smaller or comparable percentage of errors with a much smaller training sample than neural networks with certain architectures.
Highly reliable two-factor biometric authentication based on handwritten and voice passwords using flexible neural networks
hybrid networks, quadratic forms, Bayesian functionals, handwritten passwords, voice parameters, wide neural networks, biometrics-code converters, protected neural containers.
http://www.mathnet.ru/links/c753a0abcdeab6ac5da8553580c3680d/co765.pdf

Sulavko AE. Highly reliable two-factor biometric authentication based on handwritten

and voice passwords using flexible neural networks. Computer Optics 2020; 44(1): 82-91.

DOI: 10.18287/2412-6179-CO-567

The paper addresses a problem of highly reliable biometric authentication based on converters of secret biometric images into a long key or password, as well as their testing on relatively small samples (thousands of images). Static images are open, therefore with remote authentication they are of a limited trust. A process of calculating the biometric parameters of voice and handwritten passwords is described, a method for automatically generating a flexible hybrid network consisting of various types of neurons is proposed, and an absolutely stable algorithm for network learning using small samples of “Custom” (7-15 examples) is developed. A method of a trained hybrid "biometrics-code" converter based on knowledge extraction is proposed. Low values of FAR (false acceptance rate) are achieved
Identification of the Psychophysiological State of the User Based on Hidden Monitoring in Computer Systems
keystroke dynamics, biometric feature, voice attributes, identification of psychophysiological conditions, the option of working with a computer mouse
https://rdcu.be/bo0Ns

Vasiliev V.I., Sulavko A.E., Borisov R.V., Zhumazhanova S.S. Recognition of psychophysiological states of users based on hidden monitoring of actions in computer systems // Artificial intelligence and decision making. - 2017. - №. 3. - pp. 95-111.

Vasilyev V.I., Sulavko A.E., Zhumazhanova S.S., Borisov R.V. Identification of the Psychophysiological State of the User Based on Hidden Monitoring in Computer Systems // Scientific and Technical Information Processing. - December 2018, Volume 45, Issue 6, pp 398–410. doi:10.3103/S0147688218060096

It is found that the features of the voice, keystroke dynamics and pattern of a subject’s use of a computer mouse contain the following information about the psychophysiological state of the operator: normal, fatigue, intoxication, excited, and relaxed (sleepy). Voice features are the best for identifying fatigue or a sleepy state of a speaker. Keystroke dynamics, aside from these states, have features that characterize the normal state of the operator. Some features of working with a computer mouse contain information about the states of intoxication and sleepiness. This experiment on state identification was based on Bayes strategies and the neural network approach; the best result was a 5.9% error in determining the state when monitoring a subject for no more than 100 s.
Evaluation of signature verification reliability based on artificial neural networks, Bayesian multivariate functional and quadratic forms
neural networks, network of quadratic forms, multi-dimensional Bayes functional, signature reproduction peculiarities, biometric features
http://www.mathnet.ru/links/c8646912a0acff896966712654608aae/co447.pdf

Ivanov AI, Lozhnikov PS, Sulavko AE. Evaluation of signature verification reliability based on artificial neural networks, Bayesian multivariate functional and quadratic forms. Computer Optics 2017; 41(5): 765-774. DOI: 10.18287/2412-6179-2017-41-5-765-774

An experimental comparison of various functional neural networks for signature verification is performed. A signature database for the realization of the computing experiment is built. It is confirmed that up to a certain point, the increase of the decision rule dimension reduces the probability of signature verification error, with an increase in the number of neurons in the network reducing the number of errors. A higher-dimension multi-dimensional Bayes functional with stronger inter-feature correlation is found to perform better. The best result for the signature verification is obtained using networks of Bayesian multidimensional functional, with false acceptance rate of FRR = 0.0288 and false rejection rate of FAR = 0.0232.
Using networks of correlation neurons with multi-level quantization: protection against extracting knowledge from the parameters of the decision rule
One of the main global trends is related to the development of artificial intelligence (AI) technologies. Unauthorized interference in the work of AI may entail consequences (material damage, threat to life, health of citizens, violation of information security, technological failure, etc.). Therefore, in mission-critical applications, AI must run in protected mode. The report describes the results of an analytical and synthetic study of scientific works and international and national standards related to the security problems of artificial intelligence, provides a rationale for the need to protect artificial intelligence from a number of threats, and the development of a national standard “Artificial intelligence in secure execution”, and also proposes a project prospectus for this standard, which concerns only classification problems. The standard is based on the use of correlation (autocorrelation) neurons, which are a new class of neurons that analyze correlations between features instead of feature values. Networks of such neurons can be synthesized and trained automatically, while reference information about image classes is not compromised, both in execution mode and when storing the knowledge base. Networks of correlation neurons make it possible to associate a long cryptographic key with classes of images and produce it as an output when corresponding images arrive at the inputs. The potential key length exceeds all currently known methods and approaches.
Research report (part 1)

ARTIFICIAL NEURAL NETWORKS, AUTOMATIC MACHINE LEARNING, CLASSIFIER ENSEMBLES, FEATURE SPACE, ARCHITECTURAL SECURITY OF ARTIFICIAL INTELLIGENCE, ATTACKS ON ARTIFICIAL INTELLIGENCE, FEATURE EXTRACTION, RECOGNITION IMAGES
view the work on researchgate.net
Sulavko A.E., Lozhnikov P.S., Samotuga A.E. etc. Protected mode of execution of artificial intelligence based on automatically trained networks of autocorrelation neurons (part 1) // Omsk - 2021. Publishing house "OmSTU" - 101 p.
Project of the third national standard of Russia for fast automatic training of large networks of correlation neurons on small training samples of biometric data
machine learning, pattern recognition, analysis of correlations between features, meta-space of Bayes-Minkowski features, protection of confidential information in knowledge bases from compromise, secure execution of artificial intelligence, highly reliable authentication
https://cyberrus.com/wp-content/uploads/2021/04/84-93-343-21_8.-Ivanov.pdf
Ivanov A.I., Sulavko A.E. Project of the third national standard of Russia for fast automatic training of large networks of correlation neurons on small training samples of biometric data // Issues of cybersecurity. - 2021. - №. 3. - pp. 84-93. DOI:10.21681/2311-3456-2021-3-84-93

The purpose of the study is to show that a biometrics-to-access code converter based on large networks of correlation neurons makes it possible to obtain an even longer key at the output while simultaneously ensuring the protection of biometric data from compromise.
The research method is the use of large “wide” neural networks with automatic learning to implement the biometric authentication procedure while ensuring the protection of biometric personal data from compromise.
The results of the study show that the first national standard for automatic training of neural networks (GOST R 52633.5) was focused only on a physically protected, trusted computing environment. Protecting the parameters of trained neural network biometric-code converters using cryptographic methods has led to the need to use short keys and passwords for biometric-cryptographic authentication. It is proposed to build special correlation neurons in the meta-space of Bayes-Minkowski features of a higher dimension. An experiment was conducted to verify keyboard handwriting images using a biometrics-to-code converter based on the data set of the AIConstructor project. In the meta-feature space, the probability of verification error turned out to be less (EER=0.0823) than in the original feature space (EER=0.0864), while in the protected execution mode of the biometrics-code converter, the key length can be increased by more than 19 times. Experiments have shown that the transition to the Bayes-Minkowski feature space does not lead to the manifestation of the “curse of dimensionality” problem if some of the original features have a noticeable or strong cross-correlation. The problem of ensuring the confidentiality of the parameters of trained neural network containers from which a neural network biometric-code converter is formed is relevant not only for biometric authentication problems. It seems possible to develop a standard for protecting artificial intelligence based on automatically trained networks of Bayes-Minkowski correlation neurons.
On the effect of the shape of a flaw on its detectability against noise background
random fields flaw detection, human vision, computer vision, detection reliability, capability- matching systems
view the work on researchgate.net

Epifantsev B.N., Zhumazhanova S.S. On the effect of the shape of a flaw on its detectability against noise background // Russian Journal of Nondestructive Testing. 2017. Т. 53. № 1. P. 62-70

When switching to automatic output-quality testing systems, the aspiration for improving the detectability of insignificant deviations of the output parameters from the statutory ones necessitates the solution of a number of new problems. One of those is assessing the effect of the image-energy spectral density from an axisymmetric flaw on the reliability of its detection against background noise by both human and computer vision systems. Knowing this information is a necessary condition for developing new enhanced testing and evaluating techniques. Results are presented on the probabilities of false alarm and correct detection of axisymmetric circular or rectangular flaws depending on the signal-to-noise ratio (SNR) and the ratio of the flaw radius to the background-fluctuation correlation radius. It has been established that for small SNR, human vision is more effective than machine vision that implements the correlation detector algorithm and the Neyman–Pearson criterion.
Evaluation of EEG identification potential using statistical approach and convolutional neural networks
deep learning, multilayer neural networks, biometrics, machine learning, feature extraction, electrical brain activity, psychophysiological state, pattern recognition, spectrograms
http://www.i-us.ru/index.php/ius/article/view/13299
Sulavko, A. E., Lozhnikov, P. S., Choban, A. G., Stadnikov, D. G., Nigrey, A. A., Inivatov, D. P. (2020). Evaluation of EEG identification potential using statistical approach and convolutional neural networks. Information and Control Systems, (6), 37-49. https://doi.org/10.31799/1684-8853-2020-6-37-49
Introduction: Electroencephalograms contain information about the individual characteristics of the brain activities and the psychophysiological state of a subject. Purpose: To evaluate the identification potential of EEG, and to develop methods for the identification of users, their psychophysiological states and activities performed on a computer by their EEGs using convolutional neural networks.
Results: The information content of EEG rhythms was assessed from the viewpoint of the possibility to identify a person and his/her state. A high accuracy of determining the identity (98.5–99.99% for 10 electrodes, 96.47% for two electrodes Fp1 and Fp2) with a low transit time (2–2.5 s) was achieved. A significant decrease in accuracy was detected if the person was in different psychophysiological states during the training and testing. In earlier studies, this aspect was not given enough attention.
A method is proposed for increasing the robustness of personality recognition in altered psychophysiological states. An accuracy of 82–94% was achieved in recognizing states of alcohol intoxication, drowsiness or physical fatigue, and of 77.8–98.72% in recognizing the user’s activities (reading, typing or watching video).
Practical relevance: The results can be applied in security and remote monitoring applications
Using networks of correlation neurons with multi-level quantization: protection against extracting knowledge from the parameters of the decision rule
The practice of applying conventional decision rules in artificial intelligence and biometrics applications cannot guarantee their safety. It is necessary to create mathematical constructions that, on the one hand, are quite effective decision rules, and on the other hand, do not allow attackers to find out which biometric image this or that decision rule is configured for and which control action corresponds to one or another figurative situations. The Bayesian correlation neurons considered in this work are by their nature capable of providing protection against the extraction of knowledge about mathematical expectations, standard deviations and correlation coefficients controlled by biometric parameters. The article touches upon the application of Bayesian neural networks with multilevel neural network data quantizers that multiply the entropy of the internal states of the decision rule. The preprint is aimed at students, graduate students, teachers, engineers dealing with the problems of using neural network artificial intelligence to solve problems of biometrics and other applications of artificial intelligence in a research-protected execution.

protection of decisive rules from knowledge extraction attacks, neural network biometric-code converters, secure artificial intelligence
view the work on researchgate.net
Ivanov A.I., Sulavko A.E. Using networks of correlation neurons with multi-level quantization: protection against extracting knowledge from the parameters of the decision rule // Penza - 2020. Publishing house "PGU" - 48 p. Circulation 300 copies. ISBN 978-5-907364-02-8
Flexible fast learning neural networks and their application for building highly reliable biometric cryptosystems based on dynamic features
neural network, dynamic biometric features, machine learning, assessment of features quality, information security, data analysis, pattern recognition
https://www.sciencedirect.com/science/article/pii/S2405896318329458

Vasilyev V.I., Lozhnikov P.S., Sulavko A.E., Fofanov G.А., Zhumazhanova S.S. Flexible fast learning neural networks and their application for building highly reliable biometric cryptosystems based on dynamic features // IFAC-PapersOnLine. - Vol. 51, Issue 30, 2018, P. 527-532. doi:10.1016/j.ifacol.2018.11.272

The paper concludes an overview of the most promising methods of pattern recognition (deep learning, convolutional, evolutionary, progressive, shallow, "wide", hybrid artificial neural networks etc.) in regard to the possibility of their use to build highly reliable biometric cryptosystems on the basis of dynamic features. The authors propose a new approach - the development and training of flexible neural networks. For its implementation, the mathematical apparatus is developed that uses elements of various types of artificial neural networks, the probability theory, and mathematical statistics. The paper presents the results of these studies and formulates the range of the problems to be solved for creating a perspective fundamentals for building highly reliable biometric cryptosystems
Highly secure authentication using handwritten passwords based on hybrid neural networks, ensuring protection of biometric standards from compromise
pattern recognition, Bayesian difference functionals, processing of correlated biometric parameters, information security, automatic tuning of neural networks, probability densities
http://www.i-us.ru/index.php/ius/article/view/13353
Sulavko A. E. Highly reliable authentication based on handwritten passwords using hybrid neural networks with protection of biometric templates from being compromised. Informatsionno-upravliaiushchie sistemy [Information and Control Systems], 2020, no. 4, pp. xx–xx (In Russian). doi:10.31799/1684-8853-2020-4-61-77
Introduction: neural network “biometrics-code” converters are the ideological basis for a series of GOST R 52633 standards (which currently have no global analogues), which may be in demand in the development of highly reliable biometric authentication and electronic signature tools with biometric activation.
Goal: to develop a biometrics-to-code converter model for highly reliable biometric authentication using handwritten passwords with high resistance to knowledge extraction attacks.
Results: the vulnerability of neural network “biometrics-to-code” converters has been demonstrated, allowing for a quick targeted search of competing examples to compromise the biometric image and the personal key of its owner. A method of effective protection against this attack is described. A hybrid model of a “biometrics-code” neural network converter (based on a new type of hybrid neural networks) is proposed, which does not compromise the user’s biometric standard and key (password) and is resistant to such attacks. The high reliability and efficiency of the proposed model in verifying handwritten passwords have been experimentally confirmed. The reliability indicators for generating a key from a handwritten password were: FRR=11.5%, FAR=0.0009% with a key length of 1024 bits (taking into account the presentation of handwritten forgeries).
Practical significance: the results will be in demand in information security applications and in the implementation of electronic document management.
Personal Identification and the Assessment of the Psychophysiological State While Writing a Signature
biometric identification; dynamics of signature reproducing; heart rate variability; information security; user identification; psychophysiological state
https://www.mdpi.com/2078-2489/6/3/454

Lozhnikov P.S., Sulavko A.E., Samotuga A.E. Personal Identification and the Assessment of the Psychophysiological State While Writing a Signature. Information. 2015, № 6, p. 454-466. doi:10.3390/info6030454

This article discusses the problem of user identification and psychophysiological state assessment while writing a signature using a graphics tablet. The solution of the problem includes the creation of templates containing handwriting signature features simultaneously with the hidden registration of physiological parameters of a person being tested. Heart rate variability description in the different time points is used as a physiological parameter. As a result, a signature template is automatically generated for psychophysiological states of an identified person. The problem of user identification and psychophysiological state assessment is solved depending on the registered value of a physiological parameter.
Statistical approach for subject's state identification by face and neck thermograms with small training sample
artificial intelligence, applications, intelligent systems, applications, human aspects of safety, risk engineering
view the work at reader.elsevier.com
Zumazhanova, S.S., Sulavko, A.E., Ponomarev, D.B., Pasenchuk, V.A. Statistical approach for subject's state identification by face and neck thermograms with small training sample // IFAC-PapersOnLine. - Vol. 52. - Issue 25 (2019) 46–51
Existing methods of human psychophysiological states identification are either contact, or do not allow a test without the active involvement of a subject in this procedure. The use of thermal imaging systems has number of advantages: the lack of physical contact with the system, and the fact that temperature is a reliable indicator of health. In the work a new feature space consisting of 465 features divided into static (mean values in each face and neck area, correlation of temperatures between areas in each frame) and dynamic (ratio or correlation of temperatures between areas for subsequent frames) groups is proposed. The evaluation of the informativeness of the obtained features for psychophysiological state recognition is provided. The most informative features were found. Several classifiers were tested: a perceptron-type neural network, a network of quadratic forms, as well as a modified Bayes hypothesis formula, which showed the best result. The predicted probability of errors is less than 0.005.
Biometric protection of hybrid document flow
The monograph outlines a range of current problems for mixed document flow protection systems. It is proposed to move to the concept of hybrid document flow, the key difference of which is the use of biometric characteristics in the formation of a private (secret) electronic signature (ES) key. A model and technology for protecting hybrid document flow based on biometric data of handwritten images, keyboard handwriting, and face have been developed. The indicated features are analyzed, and their informativeness is assessed. Modern algorithms for forming decisions in recognizing subjects and generating key sequences based on biometric data (fuzzy extractors, neural network biometric-code converters based on perceptrons, a learning algorithm according to GOST R 52633.5-2011, networks of quadratic forms, multidimensional Bayesian functionals, and other functionals) are considered. Optimal algorithms for solving the assigned problems have been determined. A number of important theses have been identified and experimentally confirmed.

authentication, electronic signature with biometric activation, handwritten signature, biometric image, biometric code converters, correlation between external ones, mixed document flow, information security, neural networks, Bayes formula
https://kasib.ru/wp-content/uploads/2017/12/hybridocsign.pdf
P.S. Lozhnikov. Biometric protection of hybrid document flow: monograph / Novosibirsk: Publishing House SB RAS, 2017. - 130 p.
Subjects Authentication Based on Secret Biometric Patterns Using Wavelet Analysis and Flexible Neural Networks
correlation between biometric parameters, quality of features, errors of 1st and 2nd kind, wide neural networks, cybersecurity
view the work on researchgate.net

Sulavko A.E., Volkov D.A., Zhumazhanova S.S., Borisov R.V. Subjects Authentication Based on Secret Biometric Patterns Using Wavelet Analysis and Flexible Neural Networks // 2018 XIV International Scientific-Technical Conference on Actual Problems of Electronics Instrument Engineering (APEIE). - Novosibirsk, Russia. IEEE. - October 2, 2018. - P. 218-227. doi:10.1109/APEIE.2018.8545676

The article investigates the problem of reliable authentication tools construction based on secret dynamic biometric patterns of a subject. Developing such tools, it is proposed to use wavelet analysis of keystroke dynamics, handwritten and voice passwords, as well as hybrid neural networks that adapt to the characteristics of the biometric pattern of a person. The estimation of informativeness of the features, obtained using wavelet decomposition of dynamic biometric patterns with Haar, Daubechies, Morlet, Mayer, Shannon, "Mexican hat" and others bases, is given. The reliability of subjects recognition by features of voice, keyboard and handwriting in the process of secret and open biometric patterns reproduction is provided.
Bayes-Minkowski measure and building on its basis immune machine learning algorithms for biometric facial identification
pattern recognition, feature space, biometrics, correlation dependence of features, informativeness of features, robustness of machine learning
https://iopscience.iop.org/article/10.1088/1742-6596/1546/1/012103/pdf

Sulavko A.E. Bayes-Minkowski measure and building on its basis immune machine learning algorithms for biometric facial identification // Journal of Physics: Conference Series. - Vol. 1546. - IV International Scientific and Technical Conference "Mechanical Science and Technology Update" (MSTU-2020) 17-19 March, 2020, Omsk, Russian Federation. - doi:10.1088/1742-6596/1546/1/012103

In this paper we propose a new Bayes-Minkowski proximity measure that can be used to process correlated biometric, biomedical, and other type of data (with the normal distribution law or close to it). The Bayes-Minkowski measure is an antagonist criterion with respect to the Minkowski measure, since it shows opposite properties. It is possible to build a hybrid network of classifiers and apply immune learning algorithms to the network based on these proximity measures. It was demonstrated in the work on the example of tasks of identification and verification of a person’s personality by facial image. The achieved errors probabilities of person’s identification and verification by face features were: 0029 и 0.0017, respectively.
Identification of electroencephalogram images of computer system users when typing passphrases on the keyboard
electroencephalogram parameters, pattern recognition, keystroke dynamics, biometric identification, feature space, machine learning, information security.
view the work on researchgate.net

Sulavko A.E., Zhumazhanova S.S., Stadnikov D.G. Identification of electroencephalogram images of computer system users when typing passphrases on the keyboard // Artificial intelligence and decision making. №. 2. - 2019. - pp. 15-27. DOI 10.14357/20718594190202

The article discusses the relationship of the keyboard handwriting of a computer user and the parameters of his electroencephalogram (EEG). As part of the work, an experiment was conducted to collect EEG data of 65 subjects who entered the passphrase on different keyboards at different times. An EEG analysis was performed, patterns and EEG parameters (features) were identified that can be used for person biometric identification. A method for identifying a person by the characteristics of the EEG in the process of keyboard input is proposed. A computational experiment was conducted with a large volume of test sample to assess the reliability of recognition of subjects. According to the results of the experiment, 1.62% errors were obtained. At the same time, no dependence of the EEG signs on the keyboard used by the subjects and the time of day, as well as on the variability of the signs with time, was detected.
Biometric authentication on the basis of electroencephalograms parameters
neural interfaces, EEG, analysis of biomedical signals, machine learning, Bayes classifier, neural network algorithms, information security, biometrics, authentication
https://iopscience.iop.org/article/10.1088/1742-6596/1260/2/022011/pdf

Sulavko A.E., Samotuga A.E., Stadnikov D.G., Pasenchuk V.A., Zhumazhanova S.S. Biometric authentication on the basis of lectroencephalograms parameters // IOP Conf. Series: Journal of Physics: Conf. Series. III International scientific conference "Mechanical Science and Technology Update", 23-24 April, 2019. Omsk, Russia. p. 022011 doi:10.1088/1742-6596/1260/2/022011

Static biometric patterns such as fingerprint, iris and face are difficult to keep secret. Since the open pattern has a little potential replacement options stealing a strange open biometrics provides great opportunities for compromising systems. Authentication on the basis of electroencephalogram pattern (EEG) is the most secure kind of biometric security. The present study aims to develop a method of biometric authentication by the EEG data with high accuracy. Several neural network EEG pattern verification algorithms have been tested. A method for verification of the human EEG pattern based on a modified Bayes hypothesis formula has been developed. The following error indicators FAR <10-4 with FRR = 0.062 were achieved.
Identification potential of computer system users in the process of their professional activities
A range of new topical problems are outlined, the solution of which is possible using the technology of hidden recognition of a human operator and his psychophysiological state based on dynamic biometric characteristics in the process of professional activity. The state of this problem is analyzed in patent and periodical literature, and directions for further research are formulated. Approaches to solving the formulated problems are proposed based on the analysis of the parameters of handwritten images, keyboard handwriting, as well as thermograms of the subject’s face and neck. The author's methods for constructing the feature space have been tested. Various approaches to decision-making in the space of low-informative biometric features are considered. Methods have been developed for recognizing subjects and their psychophysiological states using Bayes-Hamming networks and other functionalities. The test results of the proposed systems are presented. It is recommended for graduate students and specialists whose area of interest is related to the creation of artificial intelligence systems in the information, computing, and transport fields.

image recognition, biometric technologies, operator state monitoring, Bayesian strategy, face recognition, keyboard handwriting, signature dynamics, machine learning in information security systems
https://www.elibrary.ru/item.asp?id=32257797
Epifantsev, B.N. Identification potential of users of computer systems in the process of their professional activities [Electronic resource]: monograph / B.N. Epifantsev, A.E. Sulavko, A.S. Kovalchuk, N.N. Nigray, S.S. Zhumazhanova, R.V. Borisov. – Omsk: SibADI, 2017. - 1 electron. wholesale disc (DVD-R). - Cap. from the disc label.
Alternative scenarios of authorization at identification of users by dynamics of subconscious movements
biometric identification, image recognition, authorization, deception systems, methods of integration
https://www.elibrary.ru/item.asp?id=19047630

Epifantsev B.N., Lozhnikov P.S., Sulavko A.E. Alternative authorization scenarios for identifying users by the dynamics of subconscious movements // Issues of information security / FSUE "VIMI". - Moscow: 2013, №. 2. pp. 28-35.

Proposed the concept of alternative scenarios of authorization for access control systems. Proposed the algorithm of identification of users on the dynamics of unconscious movements with alternative scenarios authorization. Conducted the experiment to evaluate the effectiveness of the algorithm.
Biometric authentication by keyboard handwriting with force of pressing the keys, parameters of vibration and movements of the operator's hands
dynamics of hand movements over the keyboard, pressure on the keys, key holding time, pauses between keystrokes, “wide” neural networks, wavelet analysis, amplitude spectrum, fast Fourier transform
view the work on researchgate.net

Sulavko, A. E. Biometric authentication by keyboard handwriting with force of pressing the keys, parameters of vibration and movements of the operator's hands / A. E. Sulavko, A. R. Khamzin, A. A. Lyzhin, M. D. Novikov, N V. Sednev, S. V. Khabarov // Issues of information protection. - 2018. - №. 2. - pp. 41-50

The paper deals with the problem of biometric authentication using the keyboard handwriting. Traditional signs of the keyboard handwriting are of little informative and do not allow creating reliable authentication means. In this paper it is suggested to use additional features: the force of pressing the keys, the trajectory of the movement of hands over the keyboard and the parameters of its vibration when typing a passphrase. To register new features, a special keyboard has been developed. An estimation of the information content of these characteristics was carried out. It is proposed to use flexible hybrid neural networks, capable of rapid learning, for recognizing keyboard users. The reliability of network decisions is estimated. The achieved result exceeds those obtained earlier.

Made on
Tilda