Computer Science and Cybersecurity

EVALUATION OF THE RESULTS OF SPATIAL CONVERSIONS OF BASIC BLOCKS OF CONTENT AS A SEPARATE STAGE OF A HYBRID STEGANOGRAPHIC ALGORITHM

Mykyta Honcharov, Serhii Malakhov — Tue, 30 Dec 2025 00:00:00 +0000

The work presents the results of modeling and analyzing the consequences of implementing procedures for changing the spatial orientation of basic blocks (BBs) of images extracted from an array of lengths series of BBs content. The purpose of this work is to determine the character of the influence and consequences of using various variants of spatial conversion of BBs of an image-content on its resistance to attacks and computational complexity. The conducted modeling demonstrated that spatial transformations of BBs of content, despite their low computational complexity and complete reversibility, provide an effective and independent level of protection. Integrating the appropriate procedures, in compatible with other levels (tools) of protection, significantly enhances the final effect, improving the robustness of steganographic content against attempts at unauthorized extraction. The evaluation of the computational complexity of spatial conversions of BBs of content confirmed the possibility of ensuring resource consensus when performing these procedures, even in conditions of a shortage of free resources used by hardware platforms. It was concluded that the wide combinatoriality of possible schemes for implementing spatial transformations of BBs is an effective and computationally «light» tool for countering attempts at unauthorized access to content. The results obtained confirm the prospects for applying the mechanism of changing the spatial orientation of BBs of content in low-resource algorithms for steganographic information protection and/or corresponding mobile applications. This opens up broad opportunities for further improvement of the considered concept of steganographic insertion image by expanding the combinatorics of the structure of the data extractor key, adaptive selection of processing parameters, and combining variants of spatial transformations of BBs of image content.

THE CONCEPT OF AN INTELLIGENT INFORMATION SYSTEM FOR CONDUCTING ACCEPTANCE TESTING OF DEEP LEARNING NEURAL NETWORKS

Yurii Halaichuk, Maryna Miroshnyk, Elvira Kulak — Tue, 30 Dec 2025 00:00:00 +0000

In the modern world, an increasing number of critical infrastructures and commercial systems rely on the results of computations by artificial intelligence algorithms, particularly neural networks. In parallel, the process of evaluating the quality of these algorithms and ensuring proper execution of all stages of their testing has become highly significant to eliminate potential flaws and ensure their ability to deliver expected results. The article addresses the issue of improving the User Acceptance Testing (UAT) process for domain-specific software utilizing deep learning neural networks. It examines challenges related to limited resources, insufficient UAT team expertise in machine learning, and the complexity of testing systems that continue learning post-initial development. A general overview of existing solutions is provided, highlighting their advantages and drawbacks. A concept of an intelligent information system based on a predictive model for evaluating neural network quality metrics, specifically accuracy and loss function is proposed, enabling the
quality assessment process of such networks using a set of training and validation data. An experimental methodology is described, including the algorithm of development of a predictive model for analyzing network quality trends and the creation of an intelligent information system to streamline and accelerate the UAT process. The system’s component deployment architecture is presented, covering interactions between client applications, a web server, an execution server, and a database, leveraging modern network protocols and technologies. The research results aim to enhance UAT efficiency through automation and the application of a predictive model to obtain dynamic quality metrics for deep learning neural network algorithms.

6G TECHNOLOGY: THE TIME HAS COME

Mhnd Farhan — Tue, 30 Dec 2025 00:00:00 +0000

The sixth generation of wireless communication technology, or 6G technology, was created to replace 5G. Compared to its predecessors, it promises much faster speeds, more capacity, and reduced latency, opening up new applications and advancing a number of industries. Terabits per second (Tbps) is the target data rate for 6G, which is substantially faster than 5G's gigabits per second (Gbps). In order to facilitate real-time applications and instantaneous data transfer, 6G aims for nearly zero latency, possibly as low as the microsecond level. Compared to 5G's 1 million connected devices per square kilometer, 6G will allow for a potentially 10 million more. With the help of AI and machine learning, 6G will be able to manage resources intelligently, perform better, and add new features. It is anticipated that 6G will facilitate developments in fields such as imaging, location awareness, presence technology, and the Internet of Things (IoT). A review of earlier work is presented in this paper.

METADATA ANALYSIS OF ENCRYPTED TRAFFIC TO ELIMINATE SECURITY «BLIND SPOTS» OF MODERN INFORMATION SYSTEMS

Maksym Horelko, Serhii Malakhov — Tue, 30 Dec 2025 00:00:00 +0000

A review of recent developments is offered on the issues in the complex analysis of encrypted network traffic in modern information systems. The main research methods are: - analysis, generalization and comparison. The paper considers the issue of finding possible ways to ensure a compromise in the conditional triangle of «influence factors» when solving the tasks of operational detection of dangers in the data structure of encrypted traffic. As «influence factors» a combination of the following factors is considered: - the need to ensure the required level of Information Security (IS); - support for the right of users to their confidentiality; - resource consensus of the implemented software and hardware solutions. Attention is drawn to the fact that the integration of artificial intelligence and machine learning (AI/ML) technologies into the structure of network traffic control algorithms is a key lever for influencing the final result. It is emphasized that the opposing party will also use these technologies to mask its activities. It is concluded that the implementation of procedures for analyzing network traffic metadata is a compromise solution. The implementation of such an approach allows to improve the «transparency» of current network activity for early detection of security threats, without directly resorting to traffic decryption procedures. It is emphasized that the implementation of the «Cyber Deception» paradigm and a comprehensive analysis of the metadata of circulating encrypted traffic are a promising vector of efforts for preventive elimination of the prerequisites of the formation of "blind spots" in the security of modern IT systems.

DEVELOPMENT AND SOFTWARE IMPLEMENTATION OF A RAILWAY ROUTING MODEL

Arte Panchenko, Iryna Zaretska, Maryna Vladimirova, Alina Biletska — Tue, 30 Dec 2025 00:00:00 +0000

Under current operating conditions of Ukrzaliznytsia, routing strategy and route selection play a key role in ensuring the economic efficiency of freight transportation. At present, routes are predominantly selected according to the shortest-path principle, which minimizes fuel consumption for transportation and reduces the wear of traction units and other rolling stock. Consequently, customers pay the lowest possible delivery cost. However, such routes are typically fixed for a predefined period and do not support dynamic adjustment, which leads to several issues, including the failure to account for the current technical condition of rolling stock on specific sections and their actual congestion levels during transportation. These problems have become particularly acute as a result of the full-scale invasion of Ukraine by the Russian Federation, which caused the destruction of parts of the railway infrastructure, including tracks and bridges, as well as damage to or complete loss of part of the traction rolling stock of Ukrzaliznytsia and other railway operators. At the same time, alternative routing approaches remain insufficiently explored due to the limited number of studies devoted to strategic management of transportation processes. This paper describes the development and implementation of a software model of railway system operations aimed at enabling experimental investigation of various hypotheses regarding alternative routing approaches. This provides a basis for solving a scientific and applied problem of optimizing freight rail transportation through the design of flexible management strategies. The study is based on a synthesis of graph theory (representation of the network as a weighted multidigraph), discrete-event simulation (DES) for analyzing process dynamics, and mixed-integer linear programming (MILP) for generating benchmark performance indicators. A hybrid threshold-based dispatching policy is implemented, relying on parameters of minimum train fill level and maximum waiting time, thereby balancing node capacity utilization and delivery times. A specialized simulation framework has been developed in Python that integrates the event lifecycle (Spawn, Form, Depart, Arrive) and enables testing of intelligent control strategies in a simulated environment. The practical significance of the research lies in the possibility of using the developed toolkit for quantitative evaluation of different routing and train formation strategies at classification yards. The created software complex serves as a fundamental platform for further research aimed at minimizing average rolling stock turnaround time and node-related dispatch delays in real logistics systems, thereby improving the economic efficiency of freight transportation.