articles
|
№ 4(76)
31 august 2018 year
Rubric: Data protection Authors: Barmina S., Tadzhibaeva F., Tumbinskaya M. V. |
DDoS-attacks are included in the top-10 network attacks and lead to serious failures in the work of
web-resources. In the paper DDoS-attacks, their classification and methods of protection are considered.
Particular attention is paid to the most common type of DDoS-attacks ‒ SYN-flood attacks, correlation
analysis of their time series and forecasting. In the paper, the correlation analysis of the time series of SYNflood
attacks is implemented, the coefficient of data autocorrelation and seasonal indices are determined.
The forecasting of SYN-flood attacks for the coming quarters of 2018 and 2019 was carried out using
simple exponential smoothing. During the investigation, it was revealed that protection against DDoSattacks
should be implemented taking into account the mechanism of DDoS-attack. Depending on it, you
should use hardware, software or mixed protection. Investigation of SYN-flood attacks showed that this
type of attacks is the most dangerous: acting at the network level and clogging the network channel with
parasitic traffic, SYN-flood attack destroys the end equipment. The correlation analysis of the time series
of SYN-flood attacks showed that they are seasonally characterized: the greatest number of attacks is expected
in the first and third quarters of 2018 and 2019. For DDoS-attacks lasting up to 4 hours, seasonality
in the first quarter of the calendar year was also revealed, which means that in the first quarter of 2019
the greatest number of attacks of a given duration should be expected. The relation between SYN-flood
attacks and attacks of up to 4 hours is moderately strong, not critical, can be traced when the delay is 3
lag time (in our case, 3 quarters). According to the forecast made using the simple exponential smoothing
method, 57.1% of DDoS attacks of the SYN type are expected in the second quarter of 2018, in the
III and IV quarters of 2018 their number will be respectively 55.35% and 57.12%. In the first quarter of
2019, 58.73% of SYN-flood attacks are expected, in the II quarter of 2019 ‒ 57.08%.
Continue...
|
|---|---|
|
Classification problems represent the group of machine learning methods where each instance is
associated with a certain category or label. An individual classifier like Neural Networks, or Decision
Trees is conventionally trained on a pre-marked or processed data set. Depending on the parameters
distributions the data sets may feature issues when all the indicators are not learned efficiently by such
a classifier, and this results in an inconsistent performance on the test sets. Ensemble classifiers denote
a set of individual classifiers algorithms that are simultaneously trained in a classification problem. The
paper aim is twofold. We present an ensemble of classifiers approach with a high predictive power for
the Russian trade-related companies bankruptcy prediction. At the first stage we split the data into a
train set (70%) and a test set (30%). At the second stage the precision of standard algorithms is measured
as applied to the empirical indicators of the data. The algorithms are trained and tested, and then
compared via the performance metrics. The standard algorithms include: random forest, decision trees
and the modifications: the chi-square automatic interaction detection (CHAID), classification and regression
trees (CRT, C5), Quick, Unbiased, Efficient, Statistical Tree (QUEST), discriminant analysis
LDA, support vector algorithms (LSVM, SVM), neural networks (multilayer and radial). Based on the
ROC-curve metrics and the prediction ability of the algorithms we select the most efficient methods
that form the ensemble of classifiers algorithm. The empirical data set included 713 trade companies
(334 — known bankrupts). The results feature the efficiency of the ensemble of classifiers algorithms
based on the simple voting (the precision metric outperforms the one of the other individual algorithms,
e.g. random forest, SVM, Logit). We also show that including the macroeconomic factors improves
the prediction power of almost all studied algorithms by at least 8%. Given that, more sophisticated
variations of the classifiers such as multilayer neural networks and random forests demonstrate higher
precision and recall with the external variables employed in the training process.
Continue...
|
|
|
№ 4(76)
31 august 2018 year
Rubric: Models and Methods Authors: Dli M. I., Bulygina O. V., Kozlov P. |
The analysis of electronic text documents written in natural language is one of the most important
tasks implementing in systems of automated analyzing linguistic information. Today the most complicated
problem is analyzing unstructured text documents coming to various organizations and authorities
through the electronic communications. The increasing volume of such documents leads to
the need to rubricate incoming messages, i.e. to solve the classification task.
The analysis of the scientific works in this field has showed the impossibility of constructing a unified
model for rubricating unstructured electronic text documents in various situations. The main reasons
are the lack of statistical data, the dynamism of the thesaurus and the small size of the incoming
document.
To solve this problem, we propose a multimodel approach to the rubrication that is characterized
by the combined use of intellectual and probabilistic-statistical methods of the text document analysis.
The choice of a specific model is carried out using fuzzy logic algorithms based on the proposed
characteristics (the size of document, the degree of rubric thesaurus intersection, the frequency of
meaningful keywords, etc.).
The implementation of the proposed multimodel approach will improve the accuracy of attributing
unstructured electronic text documents to concrete rubrics taking into account their specificity and
various objectives of practical application in the organization.
Continue...
|
|
The development of information technologies and it infrastructure is radically changing society,
management and public administration. However, there is no clear for all stakeholders common system
of concepts in the field of it, providing mutual understanding between them, a coherent view of the
changes. Therefore, it is necessary to build a basic system of concepts that allows to structure and analyze
the changes in a complex way. The article considers such a system of concepts, including terms:
conditions, object, subject, situation, object activity and others in connection with the trend of virtualization
of the properties of the object environment surrounding a person. This allows us to structure
the technical and socio-economic phenomena described by the metaphors of the Digital economy, the
Internet of things, Industry 4.0 and others. The proposed approach will provide structural completeness
of the description of situations, contribute to the solution of problems of semantic interoperability
and goal-setting. Such an approach can be applied in the process of analyzing the functionality of
information technologies, in the development and support of information technologies throughout the
life cycle, can be a conceptual basis for the development of new information technologies and in the
search for their new applications.
Continue...
|
|
|
Practice shows that at the same time, as traditional retail sales decline, the e-commerce market has
been growing steadily since 2007. The efficiency of this distribution channel depends on the chain of
connected business-processes of delivery from the supplier to the final buyer. Without proper coordination
of these processes, efficient retailing is difficult. The retail trade procedures are based on the
integrated interaction of information technology, large data technologies, Internet Things and newly
emerging technologies. This article provides an overview of the e-commerce market, the role of IT
in it, the practice of using IT in the channels for the provision and distribution of goods, and the actual
requirements of customers for the services of the reseller. On the basis of the obtained data on the
current practices of IT architecture management, it is necessary to develop reference models for retail
enterprises.
At the moment, there are developed reference models of companies in the field of retail sales, taking
into account the use of the mentioned information technologies and services, such as IoT and Big
Data. After analyzing the results of the work done, it is possible to develop reference architecture
models using modern Internet technologies Things and Big Data based on the development of integration
platforms and applicable to any information systems and services providers based on the best
retail practices.
Continue...
|
|
|
New tasks of human resource management require the analysis of huge volumes of semistructured
text information. Methods of text processing and machine learning can significantly
improve its effectiveness in case they take into consideration the features of tasks to be solved.
The article describes actual analytical problems of human resource management, characteristics
of information support of these problems, shortcomings and assumptions of frequently used
methods of both classes in the tasks context. An example of applying test processing and
machine learning methods in the task of compliance assessment is given in the article as well.
Continue...
|
