In this work, it is shown that, based on the models of artificial immune networks proposed by the authors, it is potentially possible to build AI systems with similar properties. The work does not consider all AI tasks, but only a narrow range of tasks related to biometric identification and authentication.
To read the full-text of this research,
you can request a copy directly from the authors.
ResearchGate has not been able to resolve any citations for this publication.
This work presents a method based on information-theoretic analysis of iris biometric that aims to extract homogeneous regions of high entropy. Successful extraction of these regions facilitates the development of effective systems for generation of cryptographic keys of lengths up to 400 bits per iris. At the same time, this approach allows for the application of simpler error correction codes with equal False Accept Rate levels, which reduces the overall complexity of this class of systems. The problem of neural network structure selection is discussed. The constructive approach to shallow NNs design based on using the minimal complexity principle is offered. The peculiarities of its application to solving different classes of tasks often met with in practice are considered. We provide formal definitions and efficient secure techniques for turning noisy information into keys usable for any cryptographic application, and, in particular, reliably and securely authenticating biometric data. Our techniques apply not just to biometric information, but to any-keying material that, unlike traditional cryptographic keys, is (1) not reproducible precisely and (2) not distributed uniformly. We propose two primitives: a fuzzy extractor reliably extracts nearly uniform randomness R from its input; the extraction is error-tolerant in the sense that R will be the same even if the input changes, as long as it remains reasonably close to the original. Thus, R can be used as a key in a cryptographic application. A secure sketch produces public information about its input w that does not reveal w and yet allows exact recovery of w given another value that is close to w. Thus, it can be used to reliably reproduce error-prone biometric inputs without incurring the security risk inherent in storing them. We define the primitives to be both formally secure and versatile, generalizing much prior work. In addition, we provide nearly optimal constructions of both primitives for various measures of "closeness" of input data, such as Hamming distance, edit distance, and set difference.
Технология использования больших нейронных сетей для преобразования нечетких биометрических данных в код ключа доступа
Б С Ахметов А И Иванов В А Фунтиков А В Безяев Е А Малыгина
Ахметов Б. С., Иванов А. И., Фунтиков В. А., Безяев А. В., Малыгина Е. А. Технология использования больших
нейронных сетей для преобразования нечетких биометрических данных в код ключа доступа. -Алматы: ТОО "Издательство LEM", 2014. -144 с.
Высоконадежная двухфакторная биометрическая аутентификация по рукописным и голосовым паролям на основе гибких нейронных сетей // Компьютерная
Сулавко А. Е. Высоконадежная двухфакторная биометрическая аутентификация по рукописным и голосовым
паролям на основе гибких нейронных сетей // Компьютерная
оптика. 2020. Т. 44. № 1. С. 82-91. DOI: 10.18287/2412-6179-CO-567.
Биометрическая аутентификация пользователей информационных систем по клавиатурному почерку на основе иммунных сетевых алгоритмов // Прикладная информатика
Сулавко А. Е., Шалина Е. В. Биометрическая аутентификация пользователей информационных систем по клавиатурному почерку на основе иммунных сетевых алгоритмов //
Прикладная информатика. 2019. № 3 (81). С. 39-53.
Нейросетевая защита конфиденциальных биометрических образов гражданина и его личных криптографических ключей
Иванов А. И. Нейросетевая защита конфиденциальных
биометрических образов гражданина и его личных криптографических ключей. -Пенза, 2014. -57 с.
Протоколы биометрикокриптографического рукопожатия // Системы безопасности
Иванов А. И., Чернов П. C. Протоколы биометрикокриптографического рукопожатия // Системы безопасности.
2018. № 6. С. 2-7.
Usage of fuzzy extractors in a handwritten-signature based technology of protecting a hybrid document management system
P S Lozhnikov A E Sulavko D A Volkov
V. 6. № 2. P. 89-96. DOI: 10.1049/iet-bmt.2016.0061.
11. Lozhnikov P. S., Sulavko A. E., Volkov D. A. Usage of
fuzzy extractors in a handwritten-signature based technology of
protecting a hybrid document management system: 2016 10th
International Conference on Application of Information and
Communication Technologies (AICT), 12-14 October, 2016. -Baku: Azerbaijan, 2016. P. 395-400. DOI:10.1109/ICAICT.
О многообразии метрик, позволяющих наблюдать реальные статистики распределения биометрических данных "нечетких экстракторов" при их защите наложением гаммы // Вестник УрФО. Безопасность в информационной сфере
А Е Сулавко А В Еременко Р В Борисов
Сулавко А. Е., Еременко А. В., Борисов Р. В. Генерация криптографических ключей на основе голосовых сообщений // Прикладная информатика. 2016. № 5. С. 76-89.
13. Иванов А. И., Сомкин С. А., Андреев Д. Ю., Малыгина Е. А. О многообразии метрик, позволяющих наблюдать
реальные статистики распределения биометрических данных
"нечетких экстракторов" при их защите наложением гаммы //
Вестник УрФО. Безопасность в информационной сфере. 2014.
№ 2 (12). С. 16-23.
Биометрическая защита гибридного документооборота
Ложников П. С. Биометрическая защита гибридного
документооборота. -Новосибирск: Изд-во СО РАН, 2017. -130 с.
Глубокое обучение. Погружение в мир нейронных сетей. -СПб: Питер
С Николенко А Кадурин Е Архангельская
Николенко С., Кадурин А., Архангельская Е. Глубокое
обучение. Погружение в мир нейронных сетей. -СПб: Питер, 2018. -480 с.
Чобан А. Г. Иммунные алгоритмы распознавания образов и их применение в биометрических системах (Обзор) // Вопросы защиты информации
А Е Сулавко Е В Шалина Д Г Стадников
Сулавко А. Е., Шалина Е. В., Стадников Д. Г., Чобан А. Г. Иммунные алгоритмы распознавания образов и их
применение в биометрических системах (Обзор) // Вопросы
защиты информации. 2019. № 1. С. 38-46.
Article Full-text available May 2022 · Wireless Communications and Mobile Computing
Through the use of blockchain technology, sensitive information may be securely communicated without the need to replicate it, which can assist in decreasing medical record mistakes and saving time by eliminating the need to duplicate information. Furthermore, the information is timestamped, which further enhances the security of the data even further. The deployment of blockchain technology in a
... [Show full abstract] range of healthcare situations may enhance the security and efficiency of payment transactions. In this way, only those who have been allowed access to patient medical information can see or modify such information. It is proposed in this study that blockchain technology be used to provide an accessible data storage and retrieval mechanism for patients and healthcare professionals in a healthcare system that is both safe and efficient. As of 1970, a variety of traditional knowledge-based approaches such as Personal Identification Recognition Number (PIRN), passwords, and other similar methods have been made available; however, many token-based approaches such as drivers’ licenses, passports, credit cards, bank accounts, ID cards, and keys have also been made available; however, they have all failed to establish a secure and reliable transaction channel. Because they are easily misplaced, stolen, or lost, they are usually unable to protect secrecy or authenticate the identity of a legitimate claimant. Aside from that, personally identifiable information such as passwords and PINs is very prone to fraud since they are easily forgotten or guessed by an imposter. Biometric identification and authentication (commonly known as biometrics) are attracting a great deal of attention these days, particularly in the realm of information security systems, due to its inherent potential and advantages over other conventional ways for identifying and authenticating. As a result of the device’s unique biological characteristics, which include features such as fingerprints, facepalms, hand geometry (including the iris), and the device’s iris, it can be used in a variety of contexts, such as consumer banking kiosks, airport security systems, international ports of entry, universities, office buildings, and forensics, to name a few. It is also used in several other contexts, including forensics and law enforcement. Consequently, every layer of the system—sensed data, computation, and processing of data, as well as the storage and administration of data—is susceptible to a broad variety of threats and weaknesses (cloud). There does not seem to be any suitable methods for dealing with the large volumes of data created by the fog computing architecture when normal data storage and security technologies are used. Because of this, the major objective of this research is to design security countermeasures against medical data mining vulnerabilities that originate from the sensing layer and data storage in the Internet of Things’ cloud database, both of which are discussed in more depth further down. A key allows for the creation of a distributed ledger database and provides an immutable security solution, transaction transparency, and the prohibition of tampering with patient information. This mechanism is particularly useful in healthcare settings, where patient information must be kept confidential. When used in a hospital environment, this method is extremely beneficial. As a result of incorporating blockchain technology into the fog paradigm, it is possible to alleviate some of the current concerns associated with latency, centralization, and scalability. View full-text Article Full-text available February 2022 · Computational Intelligence and Neuroscience
Computer vision is one of the hottest research directions in artificial intelligence at present, and its research goal is to give computers the ability to perceive and cognize their surroundings from a single image. Image recognition is an important research direction in the field of computer vision, which has important research significance and application value in industrial applications such
... [Show full abstract] as video surveillance, biometric identification, unmanned vehicles, human-computer interaction, and medical image recognition. In this article, we propose an end-to-end, pixel-to-pixel IoT-oriented fuzzy support tensor product adaptive image classification method. Considering the problem that traditional support tensor product classification methods are difficult to directly produce pixel-to-pixel classification results, the research is based on the idea of inverse convolution network design, which directly outputs dense pixel-by-pixel classification results for images to be classified of arbitrary size to achieve true end-to-end and pixel-to-pixel high-score image classification and improve the efficiency of support tensor product models for high-score image classification on a pixel-by-pixel basis. Moreover, considering that network supervised classification training using deep learning requires a large amount of labeled data as true values and obtaining a large number of labeled data sources is a difficult problem in the field of image classification, this article proposes using a large amount of unlabeled high-resolution remote sensing images for learning generic structured features through unsupervised to assist the labeled high-resolution remote sensing images for better-supervised feature extraction and classification training. By finding a balance between generic structural feature learning of images and differentiated feature learning related to the target class, the dependence of supervised classification on the number of labeled samples is reduced, and the network robustness of the support tensor product algorithm is improved under a small number of labeled training samples. View full-text Article Full-text available December 2021 · Advances in Mathematical Physics
This paper presents an in-depth study of face detection, face feature extraction, and face classification from three important components of a high-capacity face recognition system for the treatment area of hospital and a study of a high-capacity real-time face retrieval and recognition algorithm for the treatment area of hospital based on a task scheduling model. Considering the real-time nature
... [Show full abstract] of our system, our face feature extraction network is modeled by DeepID, and the network is slightly improved by introducing a central loss verification signal to train a DeepID-like network model using central loss and use it to extract face features. To further investigate and optimize the schedulability analysis problem of the directed graph real-time task model, this paper proposes a rigorous and approximate response time analysis method for the directed graph real-time task model with an arbitrary time frame. Based on the theoretical results of the greatly additive algebra, it is shown that the coherent qualifying function is linearly periodic, i.e., the function can be represented by a finite nonperiodic part and an infinitely repeated periodic part, thus calculating the coherent qualifying function independent of the magnitude of the interval time. The algorithm for high-capacity real-time face retrieval and recognition in the treatment area of hospital based on the task scheduling model is further investigated, and a face database is established by using the PCA dimensionality reduction technique. Based on the internal architecture of the processor, image preprocessing and IP core packaging are implemented, and the hardware engineering of the high-capacity real-time face recognition system for hospital visits is built using the IP-based design concept. The performance tests of the face detection model and feature extraction network show that the face detection model has a significant reduction in false-positive rate, better fitting of border regression, and improved time performance. The face feature extraction network has no overfitting, and the features are highly discriminative with small feature extraction time consumption. The high-capacity real-time face recognition system for the treatment area of hospital combined with the optimized directed graph task scheduling model can approach 25 fps, which meets the real-time requirements, and the face recognition rate surpasses that of real people. It realizes the intelligence, self-help, and autonomy of medical services and satisfies the medical needs of users in all aspects.
In the past half-century, biometric identification technology has developed rapidly because each biological individual has unique biometric points that can be measured, identified, and verified, and biometric identification technology identifies and authenticates individuals based on these unique points. Unlike traditional identification which is easily damaged, easily lost, and easily falsified, biometric identification technology which utilizes the unique characteristics of biological individuals is more convenient, fast, safe, reliable, and accurate. Biometric identification technology can be divided into two categories, one is identification by inherent characteristics of biological individuals, and the other is identification by behavioral characteristics of biologicals [1, 2]. Inherent characteristic identification currently has features like fingerprint, hand type, iris, retina, pulse, and face and has been used as biometric identification, and behavioral characteristics like voice and signature have been used as biometric identification. As a kind of biometric technology, face recognition has its unique advantages such as nonintrusiveness, convenience, friendliness, noncontact, and scalability compared with other biometric technologies in practical use scenarios, which makes it have an important position in biometric identification [3, 4]. If the time constraints of the real-time system are met, the system is said to be schedulable, that is, each task is completed before its own time limit. Therefore, it is required to be able to determine in advance whether it is schedulable . If the result is not schedulable, then you need to increase system resources or reduce task load so that the system becomes schedulable again.
Face recognition has become the current research hotspot in the field of artificial intelligence due to its huge application prospect. With the development of science and technology and the progress of society, the technical demand for conducting fast, efficient, and automatic face recognition is becoming more and more urgent. Face recognition technology first determines whether there is a face in the image or not, and if there is a face, the features of the face are extracted and matched with the known face comparison, and then, the identity of the face is recognized . The face recognition process is generally divided into four steps: face detection, face feature extraction, face matching, and recognition. Among them, face feature extraction is the most important part of face recognition, and it is good or bad directly affects the final result of our face recognition. From another perspective, face recognition is finding a way to describe faces, and this description needs to have strong interclass distinguishability and intraclass stability. There are neural network-based methods, statistical methods, geometric feature-based methods, model-based methods, and multiclassifier integration methods for face recognition, among which the vast majority of traditional face recognition feature expressions are human selected; however, in practice, human selection of features is a very time-consuming and energy-consuming matter, and the good or bad selection depends largely on experience and luck. However, with the increasing electronic, automated, and connected control systems, the functions and performance requirements undertaken by real-time systems are becoming more and more demanding. At the same time, the type, number, and a load of tasks in the system have increased dramatically, which poses a great challenge to the time verification of real-time systems . Therefore, it is of great significance to research schedulability analysis and performance optimization of complex real-time task systems.
This paper will focus on the direction of the task scheduling model for a high-capacity real-time face retrieval recognition algorithm for hospital visits. The existing algorithms for large-capacity real-time face retrieval and recognition inhospital visit sites match sparse datasets, which makes the large-capacity heterogeneous face recognition inhospital visit sites cannot achieve the performance of visible face recognition when using the same task scheduling model. At the same time, the existing face recognition task scheduling model is large and cannot meet the needs of offline inference computation of edge devices nowadays. By investigating the multitask scheduling method to make the large-capacity real-time face recognition model complete the task quickly and accurately on the embedded platform is the purpose of this paper. Chapter 1 is the introduction, which briefly introduces the background and significance of the study of the real-time face recognition system and finally summarizes the main work of this paper and the structure of the paper. The second chapter is related work, briefly describes the status of domestic and international research, and both face detection and feature extraction are obtained using task scheduling model, and finally, it also introduces how to improve task scheduling performance by adjusting hyperparameters. Chapter 3 firstly introduces the research on the recognition algorithm of large-capacity real-time face retrieval inhospital visit sites based on task scheduling model; then, we select the current multitask scheduling model with very high real time and performance, improve it, and introduce better recognition methods for large-capacity real-time face retrieval inhospital visit sites. Chapter 4 is the result analysis, through the hospital visit place high-capacity real-time face retrieval recognition algorithm scheduling analysis, hospital visit place high-capacity real-time face retrieval recognition algorithm performance analysis, and hospital visit place high-capacity real-time face retrieval recognition algorithm system analysis, which proves the better performance and efficiency of the algorithm researched in this paper, the relevant test of our face recognition system, and the analysis of our face recognition system performance and result presentation. Chapter 5 concludes with a summary of the research work in this paper, pointing out the shortcomings of the scheme, proposing corresponding improvement methods, and looking ahead.
2. Related Work
Since individual task scheduling is computationally limited, most problems require multiple task schedulers to work together to solve the target problem, hence the emergence of multitasking scheduling systems. Multitask scheduling systems have considerable robustness and adaptability and are more efficient for solving practical application problems. Addagarla et al. combined binary particle swarm optimization with contract networks and used the synergy of task scheduling to improve the accuracy of high-capacity real-time face retrieval recognition and reduce the scheduling time in the treatment area of hospital. The dynamic path planning method is accomplished by using the dynamic and reactive nature of task scheduling and adding it to the ant colony algorithm . Bartolini and Patella incorporated a microgrid composed of distributed power sources into a communication network formed by multitask scheduling technology, which enabled the system to ensure stable system power output even when the load changed drastically . The multitask scheduling technique is used to coordinate and control the large-capacity real-time face retrieval and recognition in a hospital visit site, and the multitask scheduling technique is incorporated into it to enhance the communication function of the large-capacity real-time face retrieval and recognition in a hospital visit site and to maintain a trouble-free retrieval and recognition . The information captured by task scheduling is processed by applying convolutional fusion. The task scheduling technique is incorporated into semantic Web services to make the discovery, invocation, and combination of Web services more efficient, and the multitask scheduling technique is introduced into deep reinforcement learning to solve the problem that the policy gradient increases with the number of task scheduling .
However, the rapid growth of face recognition technology in China in recent years and the booming of task scheduling have forced major research institutes, companies, and universities to invest heavily in face recognition . Choi and Cha proposed a transformation algorithm using Eigen to reach the transformation from face sketch to face photo and then perform face matching on the face photo obtained from the change . A method using locally linear embedding is proposed to transform the face sketch into a real photo image and the near-infrared light face image into a visible light face image, respectively, and then use a general method to match the face recognition . To consider the relationship between face facial image regions and their neighboring regions in a comprehensive way, Markov random field model is added to the project of heterogeneous face recognition . In this approach, since only the best candidate regions are selected for training, this leads to the deformation of facial images. Kumar et al. proposed a Markov weight field model, which can solve the problem of image deformation by selecting some selected regions to form a Markov network model . Muhammad et al. proposed a direct push learning method using Markov random fields, which combines sketch images with real photo images of the test images into the learning process, thus reducing the error of the model on the test data .
This paper proposes and studies a task model for dynamically adjusting the switching speed, which is a strict extension of the static adaptive rate-of-change task model . A schedulability analysis method was developed for such a new real-time task model. This method introduces a new directed graph task model to safely approximate the execution load of this type of rotation task. Compared with the existing static analysis methods, the proposed schedulability analysis method has higher accuracy and can effectively reduce the time complexity . Hospital visit site high-capacity real-time face retrieval recognition is a process algorithm used to describe concurrent systems, and its simple and efficient algorithmic model makes hospital visit site high-capacity real-time face retrieval recognition the basis for modeling in several fields. The CCS is promoted for high-capacity real-time face retrieval recognition at hospital sites and can be used to represent communication topologies with dynamic change capability by transmitting channels between distributed nodes along other channels, which is very powerful for formal description. Since a change in the physical location of processors in a distributed system may bring about a change in the communication topology, it is particularly suitable to describe a concurrent programming model for task scheduling using high-capacity real-time face retrieval recognition inhospital visit sites. The goal of this paper is to combine static task scheduling algorithms with multitask scheduling techniques and high-capacity real-time face retrieval recognition inhospital visits, so that task scheduling in a distributed environment can better utilize environmental resources, not only to enhance concurrency and reduce communication overhead by using multitask scheduling techniques but also to combine the simulated communication capability of high-capacity real-time face retrieval recognition inhospital visits . It defines a task scheduling model for uncertainty-oriented concurrent systems that reacts faster to network topology changes and is important for efficient utilization of resources in distributed systems.
3. Intelligent Analysis of Supply Chain Coordination Based on the Internet of Things
3.1. High-Capacity Real-Time Face Information Detection and Extraction for Hospital Consultation Places
In the face of multipose face detection task, the organization of the classifier is an important issue, especially when the features of faces in different poses show great differences, which will bring a great challenge to the detector; to solve this problem, a partitioning strategy is usually used, and the classifiers are trained separately for faces in different poses and then combined to build a multipose face detector. The partitioning strategy is the most basic strategy to deal with the multipose face detection task, but it also needs to take into account the speed and classification accuracy, and the enhancement of classification capability will bring an increase in computational cost.
We design Equation (1) by combining the localization error Dc and classification error Fc of the samples, the prediction probability of classification, and the intersection rate between the border and the actual target border and calculate a surrogate value cost for the border containing positive samples in the grid, based on which we judge whether the sample is selected to join the training.
The positioning error Dc is defined as Equation (2).
The classification error Fc is defined as Equation (3).
In Equations (2) and (3), is a certain threshold value of the intersection rate. is a coefficient that controls the sensitivity of the localization error Dc to the threshold value ; the inputs and for both costs take values between 0 and 1.
The purpose of DeepID2 for learning is to obtain the output vector, not to maximize the recognition rate; thus, the class spacing term is added to the paper. To facilitate the representation, we express the output vector obtained by DeepID2 layer extraction as a function, as in Equation (4), where denotes the extracted DeepID2 feature vector, the function denotes the feature extraction function, denotes the input face slice, and denotes the network parameters.
The function for correct classification is the objective function of softmax, which is aimed at minimizing the cross-entropy and is defined to identify the signal, as in Equation (5), where is the DeepID2 feature vector of Equation (4), denotes the classification target, denotes the softmax layer parameter, denotes the target probability distribution, and denotes the classification prediction probability distribution.
The objective function for maximizing the sample spacing is as follows, defined as the validation signal, as in Equation (6), where means that and belong to the same person, and this is when we need to minimize the interclass spacing; means that and belong to different people, and this is when we need to minimize the difference between and their distance values. The is a parameter that needs to be adjusted manually, and the purpose of proposing is that the objective function needs to be minimized, not maximized. Since the validation signal needs two samples to be calculated, the training process of the whole network needs to be changed accordingly. Two samples are randomly selected for each iteration during training and then trained. Our goal is to learn the parameter updated by stochastic gradient descent.
3.2. High-Capacity Real-Time Face Retrieval Recognition Algorithm Based on Task Scheduling Model for the Treatment Area of Hospital
The multiframe task can be described by the minimum release interval , the execution time vector , and the relative time frame . A typical example of a multiframe task model is an MPEG video stream decoding application using multiple frame types, capable of simulating the periodic arrival of video frames and different frames with different decoding times. As shown in Figure 1, a multiframe task can be represented by a graph with a chain structure (Figure 1(a)), where the solid nodes represent the entry points for task execution. In addition, the execution time of the multiframe task is variable, but the release pattern and the relative time frame are fixed (Figure 1(b)).
(a) View full-text June 2022
The need of personal recognition and identification became necessary in order to keep intruders away and to allow only the legitimate persons in order to protect privacy and to secure data from being altered. But, with the technological evolution and increasing risk of data theft, it has been a mandatory demand. A persons’s identity is not limited to his/her name, password or PIN but today, their
... [Show full abstract] biometrics are being used as an integral part of their identification. The emergence of Deep learning, a subset of Artificial Intelligence, brought a revolution in the field of computer vision and biometric identification. The Convolutional Neural Network (CNN) models of Deep learning possess a resemblance to the human brain and it brought a revolution in the field of computer vision and biometric identification. I adopted a transfer learning approach by using a pretrained CNN model, GoogleNetfor conducting my experiment. The GoogleNet model automatically does all the image processing operations as well as the extraction of features too. I chose a very unique human biometric trait for this experiment, which is a person’s Finger Knuckleprint (FKP). It bears a complex pattern and unique structures. I requested the knuckleprint samples from The Hong Kong Polytechnic University and I used its Contactless Finger Knuckle Images Database (Version 1.0). An excellent result obtained by conducting this experiment. Keyword: Finger Knuckle Print, Convolution, GoogleNet, Feature Extraction, Classifier. Read more Last Updated: 04 Jul 2022 Looking for the full-text?
You can request the full-text of this article directly from the authors on ResearchGate.