Proceeding of the Institute for Systems Analysis of the Russian Academy of Science
Scientific journal "Trudy Instituta sistemnogo analiza Rossiyskoy akademii nauk (ISA RAN)"publishes materials on a wide range of fundamental problems of developing sistems analysis methodology and its applying to solving various problems in the field of science and practice. The journal is destinet for scientists and researches working within the framework of these problems, as well as for politicians, employees of state and minicipal depaptments, specialists of enterprises and representatives of social organizations. The rules for articles submission to the journal, as well as for their rewief are given at the journal" s site
The journal is registered in Federal service of observing and controlling legislation in the area of mass media and protection of cultural heritage in 20 of May 2016. Digital certificate - PI № FS77-65718.
Journal Subscription Index in the Ural-Press e-catalog (http://ural-press.ru) is 36089.
The journal is supported by the Department of Information Technologies and Computing Systems of the Russian Academy of Sciences.
The journal is included in the list of leading peer-reviewed scientific journals recommended by the Higher Attestation Commission.
Current Issue
Vol 74, No 3 (2024)
Methods and Models of Systems Analysis
Modeling assessments of synergy effects under organizational activity in the system of natural monopoly characteristics
Abstract
Theoretical and applied approaches to assessing synergy effects under organizational activity are researched based on the models and techniques of the modern natural monopoly theory for solving problems of analysis and strategic design for infrastructure subsystems. The relationships between adequate theoretical foundation of effective market organization and the approaches to assessing natural monopoly synergy are analyzed. Within the normative identification, approaches to assessing the synergetic effects using key technological determinants/ natural monopoly activity indicators have been elaborated. The directions are determined for analyzing factors associated with assessing synergy effects within the framework of theoretical concepts of behavioral identification, and, accordingly, expanding the set of using characteristics. Some measures of state regulation related to the formation of market economic behavior of natural monopolist, primarily pricing behavior, are highlighted; these measures determine the possibility of adjusting estimates of synergy effects under organizational activity.



The action of the law of diffusion of the external environment of organizations as socio-economic systems
Abstract
Each organization can be represented as a system, the complexity of which depends on the number of elements of its components and the method of systematization. Collections of organizations form production and industry associations at the regional and national levels. The general theoretical focus on the study of organizations in economic sectors and their presentation as integral socio-economic entities makes it possible to clarify the main task of any research and practical activity of an enterprise. The article is devoted to the study of a specific organizational law - the law of diffusion of the external environment. Methods for calculating the corresponding elasticity coefficients for predicting the volume of work of organizations and their associations at any level of aggregation are considered. This category occupies a central place in the theory of this law, since it is very important for understanding the content and essence of things based on an objective interpretation of the dynamic changes in associations of organizations at various levels of management.



The problem of generating output forms of information systems documents
Abstract
The article formalizes the task of generating output (reporting) forms of documents based on dynamically changing data storage structures in information systems. An approach to automating the process of generating output forms is proposed. The data structures necessary and sufficient for representing digital data of information systems in output (reporting) documents are considered. The advantage of the proposed approach is the simplification and significant acceleration of the process of generating output forms, which is currently quite labor-intensive.



Dynamical Systems
Solving differential-algebraic equation systems with Pade approximation of matrix exponent
Abstract
A set of new numerical methods for solving linear differential-algebraic equation systems is developed. Homogenous systems can be solved, and nonhomogeneous systems with piecewise-polynomial right-hand side function. Calculation of system state at each integration step requires solving one or several (depending on the method order) systems of linear algebraic equations. The methods are based on decomposition of Pade approximation of the matrix exponent to simplest fractions. The proposed formulas make possible to avoid conversion of the differential-algebraic equation system to the ordinary differential equation system at the stage of system solving. The new methods are equivalent to some well-known Runge-Kutta type methods like Radau and Lobatto methods in terms of accuracy and steady areas. However, new methods are much more simple in theory and practical implementation, and they require several times less computational work. Methods with diagonal Pade approximations are A-stable, and methods with subdiagonal Pade approximations are L-stable. New methods can be used for solving stiff, oscillative and stiff-oscillative systems.



Text Mining
Context-independent fast text detection method for recognizing phone numbers
Abstract
Modern methods for detecting text in images are based on computationally expensive deep learning models and require a large amount of training data, including real data. In the case of text retrieval in arbitrary scenarios, the process of collecting and annotating real data for training is extremely labor-intensive and expensive due to the high variability of possible scenes. This paper presents a new method for detecting text in arbitrary images, which does not require photographs of text in real scenes to be trained and can be trained on simple synthetic data in the form of strings. The proposed neural network model is 42 times smaller than the text detector in one of the best text recognition systems in terms of quality and speed, PaddleOCR (84 KB versus 3.6 MB), which makes it an excellent choice for mobile devices. The model was tested as part of a phone number recognition system, where with its help it was possible to achieve 80.35% of correctly recognized numbers.



Community Informatics
Data as a challenge
Abstract
The paper deals with the problem of analyzing the nature of data arising at the boundary of knowledge and being. It shows the contradictory view of data science as a set of technologies and algorithms designed to solve problems of processing large amounts of poorly structured data, predicting correlations unrecognised earlier. The etymological and cognitive analysis of the transformation of the concept “data” is presented. The conceptual foundations of the formation of a new epistemological paradigm for displaying the surrounding world, which allows us to consider the problem of data in a new way.



Research of the distribution of distances between graphs with ordered vertices
Abstract
The article develops an approach based on a probabilistic model of generating objects with a distance between them. The distribution of distances between graphs with ordered vertices based on the maximum common is considered. One of the possible applications of such distances is the task of stylistic diagnostics of texts. Two algorithms for calculating distances on a set of graphs are presented. One of them consists of generating and exhaustively enumerating all pairs of trees, the second is heuristic. This is an approximate method of collecting statistics, where a given number of pairs of pseudo-random trees are iterated, since a complete search can take a long time. Using these algorithms, distance matrices between trees with a small and large number of vertices n were found. The experimental results showed that for small n the metric value does not exceed 0,5. For large n the average value of the metric grows slightly and stabilizes at the point 0,587. The hypothesis that the distribution corresponds to the normal law for n=100 was rejected using the Pearson test at a significance level of 0,1.



Pattern Recognition
Evaluation of an generalization ability of the nested contours algorithm in the mammograms analysis
Abstract
The work presents an nested contours algorithm designed for detecting pathological changes that may correspond to breast cancer on X-ray mammographic images, and provides the results of evaluating its generalization ability. This algorithm was tested on a large dataset of mammographic images with all possible variations of changes corresponding to verified breast cancer, including faintly visible and invisible ones. The overall detection accuracy of the algorithm was 90.73% for film and 96.82% for digital mammograms. A comparative analysis of using this algorithm and other modern methods of change detection on mammograms, with publicly available databases (INbreast and CBIS-DDSM), is also provided. The higher accuracy of the proposed algorithm is demonstrated. The high efficiency of detecting pathological changes, regardless of the differences in mammogram characteristics obtained from different systems, indicates the high generalization ability of the proposed algorithm.



Risk-Management and Security
Enhancing kubernetes security: the crucial role of DevSecOps
Abstract
This article highlights the significance of integrating DevSecOps (Development, security and Operations) practices into the research on detecting common attacks in Kubernetes environments. As Kubernetes gains rapid traction as a prominent container orchestration platform, the security challenges associated with containerized applications have grown in magnitude. However, traditional security methodologies often struggle to keep pace with the dynamic and fast-evolving nature of containerized environments, leaving potential vulnerabilities for malicious actors to exploit. By emphasizing the importance of DevSecOps, this article aims to underscore its role in improving the security posture of Kubernetes deployments and promoting a proactive approach to safeguarding containerized applications. The article also discusses key considerations and benefits of implementing DevSecOps in the context of Kubernetes security research.



Search for vulnerabilities in smart contracts based on machine learning
Abstract
With the rising popularity of blockchain projects, the number of decentralized applications based on them is also growing. The central element of these applications are smart contracts. This technology is still relatively new and has a number of security issues. Statistics of smart contract hacking indicate the relevance of finding vulnerabilities in smart contract code problem. The article describes 3 machine learning models for searching for vulnerabilities in smart contracts written in the Solidity language. Particular attention is paid to preparing the dataset for training and comparing it with well-known code analyzers. The metrics obtained from the results of training and testing the models suggest that the model consisting of three bidirectional recurrent BiGRU layers and three convolutional CNN layers is effective in the task of searching for smart contract vulnerabilities.


