Bulletin of V.N. Karazin Kharkiv National University, series «Mathematical modeling. Information technology. Automated control systems»

Spectral boundary value problem for coaxial shells of revolution

Vasyl Gnitko — 2023-12-11

The main objective of this study is to develop an efficient numerical approach using boundary elements to estimate natural frequencies of liquid vibrations in composite tanks. The spectral boundary value problem for liquid tanks is to find the natural frequencies and modes of free surface sloshing. The calculation of hydrodynamic forces on the walls of tanks with liquid is an important problem for ensuring the strength and stability of movement of industrial tanks and vessels. The vibrations of shell structures, including cylindrical and conical shells connected by rings, are analyzed. The area between the shells is filled with an ideal incompressible fluid. Numerical modeling uses the superposition method in combination with the boundary element method. A numerical solution of the spectral boundary value problem regarding fluid vibrations in rigid shell structures has been carried out. Frequencies and modes are determined by solving systems of singular integral equations. For the shells of revolution, these systems are simplified to one-dimensional equations, where the integrals are calculated along curves and line segments. Efficient numerical procedures are used to calculate one-dimensional integrals with logarithmic and Cauchy features. Test calculations confirm the high accuracy and efficiency of the proposed method. The importance and practical significance of the method lies in the ability to study fluid fluctuations in real compound fuel tanks of launch vehicles under different load conditions. This makes it possible to study the movement of liquid in fuel tanks and reservoirs under the action of external loads. The elaborated method will be used in computer modeling the dynamic behavior of liquid tanks and the stability study of liquid movement in compound fuel tanks of launch vehicles. In the future, it is planned to study the vibrations of elastic coaxial shells with liquid, using various composite materials.

Using anomaly detection method to detect network attacks

Svitlana Demenkova — 2023-12-11

The article is focused on the description of a model for detecting network intrusions in the network traffic based on the TCP/IP protocol stack. The main objects of a local area network have been analyzed. The main controlled parameters of each type of object have been described. The methods of anomaly detection based on both rule-based and probabilistic model analysis have been developed.

Relevance. Due to the intensive growth of information technologies and their implementation in various sectors of the municipal economy, the issue of information security becomes very relevant.

Research methods. In solving the tasks, the methods of control theory; methods of building security systems; graph theory; probability theory and mathematical statistics; methods of time series analysis; methods of predictive analytics and big data processing; methods of building high-load secure programs have been used. The main research methods used are probabilistic and verification modeling.

The results. Probabilistic and verification modeling of network attacks has confirmed the effectiveness of the proposed approach. The results of synthesis using CAD showed that the additional hardware costs do not exceed 20% compared to the standard description model.

Conclusions. The developed system model allows detection of attacks on key simulated objects. The results obtained during the tests showed a high efficiency of detecting anomalies in the numerical parameters of the model. In order to increase accuracy, it is planned to move on to the modeling of the concept of "service" and the modeling of HTTP, SMTP, and POP3 protocols. The session model also allows detecting existing and new TCP session-level attacks, as well as some types of denial-of-service attacks. The model of network traffic flows allows us to detect such types of attacks as: various types of system scanning, installation of Trojan programs (because the number of bytes in the output stream will increase), installation of ICMP shell (because the number of ICMP packets will increase). The time interval model allows detecting some types of system scanning, denial-of-service attacks, and web shell installation.

Shadow zones of an artificial neuron

Artem Novikov — 2023-12-11

The extremely widespread use of artificial neural networks in the diverse areas of application makes the study of their fundamental properties highly relevant. Such studies can be used to improve the properties of neural networks.

The key goal of the work: to determine the general properties of artificial neurons and detect the presence of zones where the field of output signals has a complex fractal structure in the space of all input signals.

Research methods: to explore the space of all input signals, a software that allows modelling the neuron's response to all possible input signals with a certain length in the given alphabet has been developed. With the help of the developed application the space of all input signals can be modulated and the field of output signals in this space is graphically determined. By using the capability of the software to change the scale of the input signal space, zones with a self-similar, fractal structure have been found.

Results: it has been established that when considering the overall arrangement of the neuron’s input signal space, specific areas – shadow zones – are present, which exhibit a complex fractal structure of output signal field. The impact of modifying theneuron’s weights and threshold on the presence and location of such zones has been established. The changes that follow an increase in the length of the input signals have been described. The fractal dimension of the structures within shadow zones has been determined.

Conclusions: the obtained general properties of neurons should significantly impact the properties of neural networks in the form of shadow zones in which the "response" of the network is extremely sensitive even to minute alterations in input signals. The presence of such zones is an extremely important factor that needs to be considered while developing neural networks.

Actual problems of building computer networks

Daniil Petrushenko — 2023-12-11

Relevance: In connection with the growing use of computers and the Internet in the modern world, the construction of computer networks is an important task. However, there are various challenges that need to be addressed in order to build efficient and secure networks.

Purpose: This article is devoted to an overview of current problems of building computer networks and methods of their solution.

Research methods: To write this article, an analysis of literary sources was carried out and the experience of practical work in building computer networks was considered.

Results: As a result of the analysis, it was found that the main problems of building computer networks are security, scalability, fault tolerance, efficiency and compliance with user requirements. To solve them, various technologies and tools for building and administering networks were considered.

Conclusions: Building computer networks is an important task in today's world, there are several problems that need to be solved to ensure efficient and secure network operation. Various methods are used to solve them, which help to improve the quality and reliability of networks. This article considered individual problems of building computer networks, as well as an example of a solution to one of them, namely, network load analysis

Keywords: computer networks, security, scalability, fault tolerance, efficiency, protection, encryption, firewalls, protocols, algorithms, duplication.

Forecasting and analytics in virtual distributed systems: Using machine learning models and analytical tools

Denys Telezhenko — 2023-12-11

This scientific article is devoted to the development of a conceptual model for the synthesis of the architecture of virtual distributed systems (VDS). The article examines key aspects of virtual distributed systems, including hardware, hypervisors, virtual machines, and management modules. The methodological principles of architecture synthesis are highlighted, starting from the analysis of system requirements, architecture design, implementation and testing, ending with the evaluation and optimization of VRS performance. The article emphasizes the importance of each stage in this process, emphasizes the need for a deep understanding of system requirements and the selection of appropriate technologies. The article pays special attention to the role of hypervisors and virtual machines in VRS, their connection to hardware and resource management capabilities. This article will be useful for virtualization and computing researchers and practitioners who are designing or optimizing virtual distributed systems. The article is devoted to important issues related to forecasting and optimization of virtual distributed systems, which is a key element of modern technological infrastructures. With the development of computing technologies and artificial intelligence, the need for effective resource management and support of high throughput in computing systems is increasing. The purpose of this scientific work is to research and analyze the application of machine learning algorithms, in particular LSTM (Long Short-Term Memory) and the attention mechanism, for forecasting and optimization of virtual distributed systems. The paper seeks to examine in detail how these technologies can improve resource management, provide higher efficiency and throughput of systems, and analyze how they help identify potential problems and optimize resource allocation based on accurate forecasts and historical data analysis. This paper uses various research methods, which include

The model of a neural network for text data censoring

Tymur Tytarenko — 2023-12-11

Relevance: Given the rapid development of Internet communications and the increasing amount of textual content, an urgent need to ensure effective censorship of textual data necessitates the relevance of this research. This is especially true for the online community, where it is essential to ensure the security and ethics of communication.

Purpose: to provide better and safer content for users who depend on reliable and secure Internet information by means of developing and implementing a neural network that will be able to identify inappropriate textual content in real time.

Research methods: methods of data processing and preparation, deep learning methods, neural network theory, artificial intelligence theory, mathematical analysis, methods of information content analysis, methods of classification quality assessment, and practical application research have been used in the course of the research. The software has been developed by using the Python language.

Results: the main achievement of the work is the development of a neural network model that censors textual information in real time, the model is highly scalable and can be trained on data from other languages.

Conclusions: The problem of text data censoring has been considered. Since this is a natural language processing task, an RNN-based neural network model, namely LSTM, has been proposed and developed. The study has shown the importance of innovative approaches in solving the problems of text data censorship, and the use of neural networks and artificial intelligence technologies is becoming a promising area for further research and implementation in this area.