As the Internet of Things (IoT) has connected large number of devices to the Internet, it is urgently needed to guarantee the low latency, security, scalable content distribution of the IoT network. The benefits of Information-Centric Networking (ICN) in terms of fast and efficient data delivery and improved reliability have raised ICN as a highly promising networking model for IoT environments. However, with the widely spread of the viruses and the explosion of kinds of network devices, the attackers can easily control the devices to form a botnet such as the Mirai. Once the devices are under control, the attackers can launch a consumer-provider collusive attack in the Information-Centric IoT context. In this attack, the malicious clients issue Interest packets that can only be satisfied by the malicious content provider, and the malicious provider replies to the clients just before exceeding the Pending Interest Table entry’s expiration time, to occupy the limited resources. In this paper, we expound the model of the consumer-provider collusive attack and analyze the negative effect of the attack. Then we propose a Reputation Value based Early Detection (RVED) mechanism to relieve the impact of the collusive attack. The method aims to adjust the packet dropping rates of different interfaces based on their reputation value, thus to protect the legitimate packets from being dropped as possible. We implement the consumer-provider collusive model and evaluate our defend mechanism in the simulator, and simulation results verify the feasibility and effectiveness against the collusive attack of the RVED mechanism.

/pdf/a-reputation-value-based-early-detection-mechanism-against-2hmzml62u7.pdf

A Reputation Value-Based Early Detection Mechanism Against the Consumer-Provider Collusive Attack in Information-Centric IoT

В данной статье решается задача построения нейронечеткой модели формирования нечетких правил и их использования для оценки состояния объектов в условиях неопределенности. Традиционные методы математической статистики или имитационного моделирования не позволяют строить адекватные модели объектов в указанных условиях. Поэтому в настоящее время решение многих задач основано на использовании технологий интеллектуального моделирования с применением методов нечеткой логики. Традиционный подход к построению нечетких систем связан с необходимостью привлечения эксперта для формулирования нечетких правил и задания используемых в них функций принадлежности. Для устранения этого недостатка актуальна автоматизация формирования нечетких правил на основе методов и алгоритмов машинного обучения. Одним из подходов к решению данной задачи является построение нечеткой нейронной сети и обучение ее на данных, характеризующих исследуемый объект. Реализация этого подхода потребовала выбора вида нечетких правил с учетом особенностей обрабатываемых данных. Кроме того, потребовалась разработка алгоритма логического вывода на правилах выбранного вида. Этапы алгоритма определяют число слоев в структуре нечеткой нейронной сети и их функциональность. Разработан алгоритм обучения нечеткой нейронной сети. После ее обучения производится формирование системы нечетко-продукционных правил. На базе разработанного математического обеспечения реализован программный комплекс. На его основе проведены исследования по оценке классифицирующей способности формируемых нечетких правил на примере анализа данных из UCI Machine Learning Repository. Результаты исследований показали, что классифицирующая способность сформированных нечетких правил не уступает по точности другим методам классификации. Кроме того, алгоритм логического вывода на нечетких правилах позволяет успешно производить классификацию при отсутствии части исходных данных. С целью апробации произведено формирование нечетких правил для решения задачи по оценке состояния водоводов в нефтяной отрасли. На основе исходных данных по 303 водоводам сформирована база из 342 нечетких правил. Их практическая апробация показала высокую эффективность в решении поставленной задачи.

/pdf/neuro-fuzzy-model-of-fuzzy-rules-formation-for-objects-state-5fodp3j3hc.pdf

Neuro-fuzzy model of fuzzy rules formation for objects state evaluation in conditions of uncertainty

The extant literature suggests that one of the main causes of the recent financial crisis has been the excessive use of short-term debt by banks [Gorton and Metrick (2012a, b)]. Using a large sample of banks we find that increases in repurchase agreements (repos) was recognized by external capital markets to increase bank risk in the pre-crisis period. In the crisis, we find a negative relationship between repos and risk. We attribute this result to evidence suggesting that “good” banks were able to continue funding their repos, whereas “bad” banks had to significantly decrease their repo funding.

Short-Term Debt and Bank Risk

Advances in Neural Information Processing Systems 20

This article considers the task of objects state assessing in conditions of uncertainty by considering the supply chain strategy. To solve it, the need to use fuzzy-production knowledge bases and fuzzy inference algorithms as part of fuzzy decision support systems is being updated. As a tool for constructing a knowledge base, a neural-fuzzy model is proposed. The proposed type of fuzzy-production rules and the logic inference algorithm on rules for objects state assessing are described. A structure of a fuzzy neural network, consisting of six layers, each of which implements the corresponding stage of the logic inference algorithm, is proposed. As a result of training a fuzzy neural network, a system of fuzzy-production rules is formed, which make up the knowledge base of the decision support system for objects state assessing. On the basis of the proposed neuro-fuzzy model, a software package has been implemented for automating the processes of forming fuzzy-production rules. The main components of the software package are the knowledge base generation module and the fuzzy inference module. As an approbation of the neuro-fuzzy model, the formation of fuzzy rules for assessing the state of water lines at the cluster pumping stations in reservoir pressure maintenance systems has been carried out. The testing results confirmed the high efficiency of the neural-fuzzy model and the possibility of its practical use for the formation of fuzzy-production rules in various subject areas of human activity.

https://ojs.excelingtech.co.uk/index.php/IJSCM/article/download/3835/1902

Neuro-Fuzzy Model in Supply Chain Management for Objects State Assessing

This paper solves the problem of defining phishing sites based on building a team model of fuzzy neural networks (FNNs). The main methods of phishing are analyzed. Attention is drawn to the fact that phishing has become widespread on the Internet through the use of phishing sites. The expediency of identifying phishing sites based on the analysis of their URLs is noted. The main approaches to identifying phishing sites are described. The need toimplement an approach based on machine learning by constructing fuzzy neural networks for the creation of fuzzy knowledge bases and their use to identify phishing sites is actualized. Automating the identification of phishing sites based on the neuro-fuzzy approach required solving the problems of collecting and preparing initial data for analysis,building a team model of fuzzy neural networks, and forming a fuzzy knowledge base, as well as conducting research, and assessing the accuracy of identifying phishing sites based on the constructed model. The initial data was formed from various sources. The total amount of initial data was 50,000. Of these, 10 input features for analysis were selected by an expert. After carrying out the correlation analysis, 4 most informative input features were selected for analysis: site lifetime, site rank, URL length, and the registered status of the site. An output feature of the site was its type: phishing or legitimate. After assessing the quality and cleaning the selected data, the resulting sample was formed of 34718 rows, of which 70% were used for learning (24303 rows), and 30% (10415 rows) for testing. A team model offuzzy neural networks was built and a knowledge base was formed on the basis of the data obtained, including 4608 fuzzy rules. Studies have shown that the number of errors of the 1st type in identifying phishing sites is 2.01%, and 2.89% for errors of the 2nd type. The general classification error based on knowledge base rules is 4.9%. The accuracy of identifying phishing sites was 95.1%, which exceeds the accuracy of other classification methods: multilayer neural network, decision tree, linear and logistic regression. The knowledge base formed on the basis of the team model of fuzzy neural networks can be effectively used to identify phishing sites on the Internet.

https://helixscientific.pub/index.php/home/article/download/237/190

Definition of Phishing Sites Based on the Team Model of Fuzzy Neural Networks

This article solves the problem of intelligent models constructing and their accuracy evaluating for identifying bots in social networks. The relevance of solving this problem is noted. The construction and accuracy assessment of the neural network model, decision tree and linear regression are performed. The initial data source was Twitter social network. To collect the initial data, we used our own database, consisted of 3428 users, about half of which contained characteristic features of bots. The initial data were randomly divided into the training and testing sets, each of them included approximately 50% of the records. 15 attributes were used as the model’s input parameters, in particular, the number of symbols in the username, the user’s number of tweets, the number of readers, etc. The models construction and study was carried out on the Deductor analytical platform base. Each model was tested on data set consisted of 1719 records. For all models, the corresponding classification matrices were constructed and the first, second kind errors and the general model’s error were calculated. In terms of minimizing these errors, the neural network model showed the best results, and the linear regression model showed the worst. This allowed us to conclude, that in order to minimize classification errors, it is advisable to use a neural network model. This indicates its effectiveness and the possibility of practical use in intelligent decision-making support systems for bots identifying in social networks.

/pdf/identification-of-bots-in-social-networks-based-on-data-58hzh8zh5h.pdf

Identification of Bots in Social Networks based on Data Mining Technologies

This article solves the problem of constructing and evaluating a neural network model to determine the creditworthiness of individuals. It is noted that the most important part of the modern retail market is consumer lending. Therefore, an adequate and high-quality assessment of the creditworthiness of an individual is a key aspect of providing credit to a potential borrower. The theoretical and practical aspects of assessing the creditworthiness of individuals are considered. To solve this problem, the need for the use of intelligent modeling technologies based on neural networks is being updated. The construction of a neural network model required the receipt of initial data on 1 Kazan Federal University. 2 Kazan Federal University. 3 Kazan National Research Technical University named after A.N. Tupolev. e-mail: molotov.linar@mail.ru. Tel.: +7 9377703937. 4 Kazan National Research Technical University named after A.N. Tupolev. e-mail: molotov.linar@mail.ru. Tel.: +7 9377703937. borrowers. Using correlation analysis, 14 input parameters were selected that most significantly affect the output. The training and test data samples were generated to build and evaluate the adequacy of the neural network model. Training and testing of the neural network model was carried out on the basis of the analytical platform “Deductor”. Analysis of contingency tables to assess the accuracy of the neural network model in the training and test samples showed positive results. The error of the first kind on the data from the training sample was 0.45%, and the error of the second kind was 1.39%. Accordingly, the error of the first kind was not observed on the data from the Periódico do Núcleo de Estudos e Pesquisas sobre Gênero e Direito Centro de Ciências Jurídicas Universidade Federal da Paraíba V. 8 No 06 Ano 2019 – Special Edition ISSN | 2179-7137 | http://periodicos.ufpb.br/ojs2/index.php/ged/index 350 test sample, and the error of the second kind was 2.68%. The results obtained indicate a high generalizing ability and adequacy of the constructed neural network, as well as the possibility of its effective practical use as part of intelligent decision support systems for granting loans to potential borrowers.

/pdf/the-neural-network-model-of-individuals-credit-rating-49rw0khpfg.pdf

Dina Vladimirovna Kataseva

Papers

Definition of Phishing Sites Based on the Team Model of Fuzzy Neural Networks

Identification of Bots in Social Networks based on Data Mining Technologies

The neural network model of individuals credit rating

Formation of knowledge bases for intellectual systems based on neuro-fuzzy analysis of medical data

Fuzzy model for objects state assessing