Yuriy Pepa
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Yuriy Pepa has a total of 55 co-authors. Here are the top co-authors by number of shared publications:
Юрій Пепа
Державний університет інформаційно-комунікаційних технологій
3 shared publications- Andrii Petrenko
National University “Kyiv Aviation Institute
2 shared publications - Oleh Deineka
Lviv Polytechnic National University
2 shared publications - Oleh Harasymchuk
Lviv Polytechnic National University
2 shared publications - Vadym Akhramovych
Національна академія статистики, обліку та аудиту, Київ
2 shared publications - Volodymyr Akhramovych
National University “Kyiv Aviation Institute
2 shared publications
Latest Publications
- Open Access
- 26 October 2026
- Published by Borys Grinchenko Kyiv University in Cybersecurity: Education, Science, Technique
- Vol. 3 (35) , 389-400
ADAPTIVE ESTIMATION AND DETECTION OF HAZARDOUS SIGNALS IN POWER SUPPLY LINES OF AN INFORMATION ACTIVITY FACILITY
The article considers the problem of synthesizing an adaptive receiver for hazardous signals in a 220 V, 50 Hz power supply line at an information activity facility (IAF). It is shown that, after filtering the power-frequency component and its harmonics, the residual interference may exhibit non-Gaussian, impulsive, and multimodal characteristics, causing the detector likelihood function to have several local maxima. A quasi-optimal approach to the joint detection and estimation of a hazardous signal is proposed. The approach is based on representing the posterior probability density as a weighted sum of partial components, forming local estimates of the information parameter, and Bayesian updating of the weighting coefficients. A structural implementation of the adaptive receiver has been developed, comprising preliminary filtering, interference environment analysis, partial processing, aggregate estimation, and decision-making units. To verify the proposed approach, simulation modeling was performed for a bimodal Gaussian mixture. For a given false-alarm probability, the proposed mixture-based detector provided the required probability of correct detection and, compared with a detector based on an equivalent Gaussian model, reduced the mean-square error in estimating the hazardous signal level by approximately four times. The obtained results confirm the feasibility of accounting for the multimodal nature of interference when conducting special investigations (SI) of wired technical information leakage channels through power supply lines at an information activity facility.
- 26 October 2026
- Published by Borys Grinchenko Kyiv University in Cybersecurity: Education, Science, Technique
- Vol. 3 (35) , 389-400
Open AccessADAPTIVE ESTIMATION AND DETECTION OF HAZARDOUS SIGNALS IN POWER SUPPLY LINES OF AN INFORMATION ACTIVITY FACILITY
The article considers the problem of synthesizing an adaptive receiver for hazardous signals in a 220 V, 50 Hz power supply line at an information activity facility (IAF). It is shown that, after filtering the power-frequency component and its harmonics, the residual interference may exhibit non-Gaussian, impulsive, and multimodal characteristics, causing the detector likelihood function to have several local maxima. A quasi-optimal approach to the joint detection and estimation of a hazardous signal is proposed. The approach is based on representing the posterior probability density as a weighted sum of partial components, forming local estimates of the information parameter, and Bayesian updating of the weighting coefficients. A structural implementation of the adaptive receiver has been developed, comprising preliminary filtering, interference environment analysis, partial processing, aggregate estimation, and decision-making units. To verify the proposed approach, simulation modeling was performed for a bimodal Gaussian mixture. For a given false-alarm probability, the proposed mixture-based detector provided the required probability of correct detection and, compared with a detector based on an equivalent Gaussian model, reduced the mean-square error in estimating the hazardous signal level by approximately four times. The obtained results confirm the feasibility of accounting for the multimodal nature of interference when conducting special investigations (SI) of wired technical information leakage channels through power supply lines at an information activity facility.
- 22 September 2026
- Published by Springer Nature
- p. 225-242
Energy-Efficient Deep Neural Networks for Sustainable and Intelligent Edge Computing
- 22 September 2026
- Published by Springer Nature
- p. 225-242
Energy-Efficient Deep Neural Networks for Sustainable and Intelligent Edge Computing
- Open Access
- 25 June 2026
- Published by Borys Grinchenko Kyiv University in Cybersecurity: Education, Science, Technique
- Vol. 1 (33) , 635-646
A METHOD FOR ADAPTIVE ESTIMATION OF THE PROBABILITY OF INFORMATION LEAKAGE IN SPECIAL-PURPOSE NETWORKS FOR MACHINE LEARNING
This article presents a formalized model of information leakage in special-purpose networks (SPN) and develops a method for adaptive estimation of the probability of information leakage in such networks. The relevance of the research is driven by the growing number of sophisticated targeted cyber threats, the use of multi-layered attacks and covert data transmission channels, as well as the increasing role of insider threats in modern information and communication systems. It is shown that traditional protection measures, focused primarily on perimeter security, signature-based detection, and static access policies, do not provide an adequate level of protection against information leakage in the context of dynamic changes in the operating environment of SPNs. This paper proposes a formalization of the SPN as a set of nodes, information transmission channels, and risk factors that underlie the occurrence of leaks. The developed method is based on the integration of network, host, and contextual security indicators, the normalization of features, the formation of a risk factor vector, the adaptive updating of weight coefficients, and the determination of an integral risk function, followed by the calculation of the probability of information leakage. A key feature of the method is its ability to be applied in real time, taking into account changes in user behavior, network operating modes, and the current threat level. An example of the practical implementation of the proposed approach is provided, demonstrating the sequence of processing security indicators, risk assessment, and decision-making regarding incident response. The results confirm that applying the proposed method improves the effectiveness of detecting precursors to information leaks, reduces the risk of compromising confidential data, and ensures that the protection system adapts to new types of threats in the context of targeted countermeasures.
- 25 June 2026
- Published by Borys Grinchenko Kyiv University in Cybersecurity: Education, Science, Technique
- Vol. 1 (33) , 635-646
Open AccessA METHOD FOR ADAPTIVE ESTIMATION OF THE PROBABILITY OF INFORMATION LEAKAGE IN SPECIAL-PURPOSE NETWORKS FOR MACHINE LEARNING
This article presents a formalized model of information leakage in special-purpose networks (SPN) and develops a method for adaptive estimation of the probability of information leakage in such networks. The relevance of the research is driven by the growing number of sophisticated targeted cyber threats, the use of multi-layered attacks and covert data transmission channels, as well as the increasing role of insider threats in modern information and communication systems. It is shown that traditional protection measures, focused primarily on perimeter security, signature-based detection, and static access policies, do not provide an adequate level of protection against information leakage in the context of dynamic changes in the operating environment of SPNs. This paper proposes a formalization of the SPN as a set of nodes, information transmission channels, and risk factors that underlie the occurrence of leaks. The developed method is based on the integration of network, host, and contextual security indicators, the normalization of features, the formation of a risk factor vector, the adaptive updating of weight coefficients, and the determination of an integral risk function, followed by the calculation of the probability of information leakage. A key feature of the method is its ability to be applied in real time, taking into account changes in user behavior, network operating modes, and the current threat level. An example of the practical implementation of the proposed approach is provided, demonstrating the sequence of processing security indicators, risk assessment, and decision-making regarding incident response. The results confirm that applying the proposed method improves the effectiveness of detecting precursors to information leaks, reduces the risk of compromising confidential data, and ensures that the protection system adapts to new types of threats in the context of targeted countermeasures.
- 9 February 2026
- Published by Springer Nature
- p. 685-701
Balancing Security and Civil Liberties: A Mixed-Methods Analysis of Constitutional Frameworks and Proportionality in Governance
- 9 February 2026
- Published by Springer Nature
- p. 685-701
Balancing Security and Civil Liberties: A Mixed-Methods Analysis of Constitutional Frameworks and Proportionality in Governance
- Open Access
- 20 December 2025
- Published by Politechnika Lubelska in Informatyka, Automatyka, Pomiary w Gospodarce i Ochronie Środowiska
- Vol. 15 (4) , 124-129
Analysis of modern tools, methods of audit and monitoring of database security
The mismatch between modern technological innovations and traditional approaches to information security can significantly affect the effectiveness of database monitoring systems. This paper examines the ethical and legal aspects of database monitoring, taking into account current regulatory requirements and the importance of maintaining data confidentiality. The study also evaluates the use of each tool to determine their effectiveness in real-world conditions, the possibility of improving the functional characteristics of monitoring systems, their adaptation to new technologies and increasing the overall level of information protection.
- 20 December 2025
- Published by Politechnika Lubelska in Informatyka, Automatyka, Pomiary w Gospodarce i Ochronie Środowiska
- Vol. 15 (4) , 124-129
Open AccessAnalysis of modern tools, methods of audit and monitoring of database security
The mismatch between modern technological innovations and traditional approaches to information security can significantly affect the effectiveness of database monitoring systems. This paper examines the ethical and legal aspects of database monitoring, taking into account current regulatory requirements and the importance of maintaining data confidentiality. The study also evaluates the use of each tool to determine their effectiveness in real-world conditions, the possibility of improving the functional characteristics of monitoring systems, their adaptation to new technologies and increasing the overall level of information protection.
- Open Access
- 30 September 2025
- Published by Politechnika Lubelska in Informatyka, Automatyka, Pomiary w Gospodarce i Ochronie Środowiska
- Vol. 15 (3) , 91-99
Detection confidential information by large language models
In today's digital age, the protection of personal and confidential customer data is paramount. With the increasing volume of data being generated and processed, organizations face significant challenges in ensuring that sensitive information is adequately protected. One of the critical steps in safeguarding this data is the detection and classification of personal and confidential information within text documents. This process involves identifying sensitive data, classifying it appropriately, and storing the results in a semi-structured format such for further analysis and action. The need for detecting and classifying sensitive data is driven by regulatory compliance, data security, risk management, and operational efficiency. Various methodologies, including rule-based systems, machine learning models, natural language processing (NLP), and hybrid approaches, are employed to detect and classify sensitive data. Large Language Models (LLMs) like GPT-3 and BERT, trained on extensive text data, are transforming data management and governance, areas crucial for SOC 2 Type 2 compliance. LLMs respond to prompts, guiding their output generation, and can automate tasks like data cataloging, enhancing data quality, ensuring data privacy, and assisting in data integration. These capabilities can support a robust data classification policy, a key requirement for SOC 2 Type 2.
- 30 September 2025
- Published by Politechnika Lubelska in Informatyka, Automatyka, Pomiary w Gospodarce i Ochronie Środowiska
- Vol. 15 (3) , 91-99
Open AccessDetection confidential information by large language models
In today's digital age, the protection of personal and confidential customer data is paramount. With the increasing volume of data being generated and processed, organizations face significant challenges in ensuring that sensitive information is adequately protected. One of the critical steps in safeguarding this data is the detection and classification of personal and confidential information within text documents. This process involves identifying sensitive data, classifying it appropriately, and storing the results in a semi-structured format such for further analysis and action. The need for detecting and classifying sensitive data is driven by regulatory compliance, data security, risk management, and operational efficiency. Various methodologies, including rule-based systems, machine learning models, natural language processing (NLP), and hybrid approaches, are employed to detect and classify sensitive data. Large Language Models (LLMs) like GPT-3 and BERT, trained on extensive text data, are transforming data management and governance, areas crucial for SOC 2 Type 2 compliance. LLMs respond to prompts, guiding their output generation, and can automate tasks like data cataloging, enhancing data quality, ensuring data privacy, and assisting in data integration. These capabilities can support a robust data classification policy, a key requirement for SOC 2 Type 2.
- Open Access
- 27 June 2025
- Published by Politechnika Lubelska in Informatyka, Automatyka, Pomiary w Gospodarce i Ochronie Środowiska
- Vol. 15 (2) , 139-146
Prediction of quality software quality indicators with applied modifications of integrated gradiates methods
The article is devoted to modern software systems (SS) and improving their quality using machine learning methods, including the Integrated Gradients (IG) method. Key problems and limitation of IG use in real operating conditions of the SS, such as complexity of systems, correlation of variables and computing efficiency are considered. Ways to improve IG, including adaptive integration, spatial smoothing and use of weight factors, are proposed. Experimental results are described that confirm the effectiveness of the proposed modifications to improve the quality of the SS. Adaptive integration has achieved the best results (MAE 0.11), adaptability and interpretation.
- 27 June 2025
- Published by Politechnika Lubelska in Informatyka, Automatyka, Pomiary w Gospodarce i Ochronie Środowiska
- Vol. 15 (2) , 139-146
Open AccessPrediction of quality software quality indicators with applied modifications of integrated gradiates methods
The article is devoted to modern software systems (SS) and improving their quality using machine learning methods, including the Integrated Gradients (IG) method. Key problems and limitation of IG use in real operating conditions of the SS, such as complexity of systems, correlation of variables and computing efficiency are considered. Ways to improve IG, including adaptive integration, spatial smoothing and use of weight factors, are proposed. Experimental results are described that confirm the effectiveness of the proposed modifications to improve the quality of the SS. Adaptive integration has achieved the best results (MAE 0.11), adaptability and interpretation.
- Open Access
- 14 May 2025
- Published by Kyiv National University of Construction and Architecture in Smart technologies: Industrial and Civil Engineering
- Vol. 3 (16) , 81-99
The method of data encryption based on the block identification technology using the matrix method
The article discusses the development and implementation of a matrix method for identifying data blocks on server platforms, which is accompanied by a constant increase in the volume of information. One of the important aspects of managing information flows is the process of identifying data blocks, which allows determining their affiliation, structure, and relationships between individual fragments of information. Since the data on the server can change, update, and duplicate, their effective identification becomes a critical task, on which the speed of query processing, the level of load on computing resources, and the overall performance of the system depend. Matrix methods allow identifying data blocks based on mathematical transformations and linear algebra algorithms. The application of matrix operations in this context helps optimize information processing processes, increase the accuracy and speed of analyzing relationships between data, and reduce computational costs. The study analyzes existing approaches and their limitations, develops a mathematical model, identification algorithm, and evaluates efficiency on test samples. The implementation of the matrix approach to identifying data blocks can find wide application in the field of database management, cloud computing, information protection, and distributed systems, confirming its relevance in the context of modern trends in information technology development.
- 14 May 2025
- Published by Kyiv National University of Construction and Architecture in Smart technologies: Industrial and Civil Engineering
- Vol. 3 (16) , 81-99
Open AccessThe method of data encryption based on the block identification technology using the matrix method
The article discusses the development and implementation of a matrix method for identifying data blocks on server platforms, which is accompanied by a constant increase in the volume of information. One of the important aspects of managing information flows is the process of identifying data blocks, which allows determining their affiliation, structure, and relationships between individual fragments of information. Since the data on the server can change, update, and duplicate, their effective identification becomes a critical task, on which the speed of query processing, the level of load on computing resources, and the overall performance of the system depend. Matrix methods allow identifying data blocks based on mathematical transformations and linear algebra algorithms. The application of matrix operations in this context helps optimize information processing processes, increase the accuracy and speed of analyzing relationships between data, and reduce computational costs. The study analyzes existing approaches and their limitations, develops a mathematical model, identification algorithm, and evaluates efficiency on test samples. The implementation of the matrix approach to identifying data blocks can find wide application in the field of database management, cloud computing, information protection, and distributed systems, confirming its relevance in the context of modern trends in information technology development.
- Open Access
- 1 January 2025
- Published by Taras Shevchenko National University of Kyiv in Information systems and technologies security
- Vol. 1 (1 (9)) , 18-25
Quantitative risk assessment using the fuzzy sets method
Background. This paper presents a quantitative risk assessment method based on the analysis and evaluation of risks in information systems. The proposed approach enables the use of a wide range of parameters, providing the creation of flexible assessment tools. This method allows calculating risks based on both statistical data and expert evaluations conducted under conditions of uncertainty and poorly formalized environments. Additionally, the developed methods provide the representation of results in both numerical and verbal forms. For example, linguistic variables, which are often used to describe complex systems characterized by both quantitative and qualitative parameters, can be utilized. The risks of information systems can be described through a conceptual fuzzy set model that accounts for uncertainty, imprecision, and subjectivity in their evaluation. Methods. The fuzzy sets method was applied to study the risks of information systems. This method facilitates the prediction of potential economic losses. The proposed approach allows the integration of various risk factors into a unified model, considering both quantitative and qualitative aspects of their assessment. Result. During the study, a quantitative approach to assessing the risks of enterprise information systems was developed, enabling a comprehensive analysis and evaluation of the impact of various risk factors. The main results of the work include: The model accounts for uncertainty, imprecision, and subjectivity of data, which is especially important in unstable environments. This ensures high adaptability of the method to different areas of enterprise activity. The proposed methodology enables the quantitative assessment of potential losses associated with risks in information systems based on statistical and expert data. Based on the risk assessment results, recommendations can be developed to improve measures for protecting the enterprise's information assets. This includes the creation of adaptive protection strategies aimed at reducing economic losses. Conclusions. The use of the fuzzy sets method for quantitative risk assessment of information systems has proven its effectiveness due to its ability to account for uncertainty and subjectivity in data evaluation. The developed approach not only predicts risks but also supports informed decision-making to reduce potential losses, contributing significantly to the development of enterprise information security management systems.
- 1 January 2025
- Published by Taras Shevchenko National University of Kyiv in Information systems and technologies security
- Vol. 1 (1 (9)) , 18-25
Open AccessQuantitative risk assessment using the fuzzy sets method
Background. This paper presents a quantitative risk assessment method based on the analysis and evaluation of risks in information systems. The proposed approach enables the use of a wide range of parameters, providing the creation of flexible assessment tools. This method allows calculating risks based on both statistical data and expert evaluations conducted under conditions of uncertainty and poorly formalized environments. Additionally, the developed methods provide the representation of results in both numerical and verbal forms. For example, linguistic variables, which are often used to describe complex systems characterized by both quantitative and qualitative parameters, can be utilized. The risks of information systems can be described through a conceptual fuzzy set model that accounts for uncertainty, imprecision, and subjectivity in their evaluation. Methods. The fuzzy sets method was applied to study the risks of information systems. This method facilitates the prediction of potential economic losses. The proposed approach allows the integration of various risk factors into a unified model, considering both quantitative and qualitative aspects of their assessment. Result. During the study, a quantitative approach to assessing the risks of enterprise information systems was developed, enabling a comprehensive analysis and evaluation of the impact of various risk factors. The main results of the work include: The model accounts for uncertainty, imprecision, and subjectivity of data, which is especially important in unstable environments. This ensures high adaptability of the method to different areas of enterprise activity. The proposed methodology enables the quantitative assessment of potential losses associated with risks in information systems based on statistical and expert data. Based on the risk assessment results, recommendations can be developed to improve measures for protecting the enterprise's information assets. This includes the creation of adaptive protection strategies aimed at reducing economic losses. Conclusions. The use of the fuzzy sets method for quantitative risk assessment of information systems has proven its effectiveness due to its ability to account for uncertainty and subjectivity in data evaluation. The developed approach not only predicts risks but also supports informed decision-making to reduce potential losses, contributing significantly to the development of enterprise information security management systems.
- Open Access
- 31 December 2024
- Published by State University "Kyiv Aviation Institute" in Ukrainian Information Security Research Journal
- Vol. 26 (2) , 270-277
СИСТЕМА РЕАЛІЗАЦІЇ ЗАХИСТУ СЕРВЕРІВ З УРАХУВАННЯМ АНОМАЛІЙ В ПАКЕТАХ
У сучасних умовах зростання кіберзагроз захист серверів стає критично важливим аспектом інформаційної безпеки, особливо в контексті збільшення обсягів мережевого трафіку та складності атак. Одним із ефективних підходів є використання систем захисту, що враховують аномалії у мережевих пакетах. Виявлення та обробка таких аномалій дозволяють оперативно ідентифікувати та нейтралізувати загрози, серед яких особливе місце займають DDoS-атаки. У статті розглянуто методи аналізу мережевого трафіку в реальному часі, що базуються на статистичних методах, а також алгоритмах машинного навчання для класифікації мережевих пакетів за їх поведінковими характеристиками [1]. Представлена система реалізує багаторівневий підхід до захисту серверів, який включає три основні етапи: первинну фільтрацію даних, статистичний аналіз та використання моделей машинного навчання. На першому етапі відсіюються шкідливі пакети на основі простих критеріїв, таких як заборонені IP-адреси або некоректний формат пакетів [2]. На другому етапі застосовується статистичний аналіз для виявлення відхилень у розподілі трафіку, наприклад, раптове збільшення кількості запитів або зміну розміру пакетів [3]. Третій етап передбачає застосування класифікаторів, які навчаються на історичних даних для визначення аномалій у поведінці мережі. Перелік представлених моделей дозволяє адаптуватися до нових типів атак шляхом автоматичного оновлення [4]. Наведемо переваги представленої системи: виявляє як традиційні DDoS-атаки (сканування портів, експлуатація вразливостей мережевих протоколів та спроби ін’єкцій SQL-запитів), так і інші види загроз. По-друге, її інтеграція з існуючими інструментами моніторингу та фаєрволами. Також забезпечує простоту впровадження без значного збільшення витрат завдяки інтеграції з існуючими інструментами моніторингу та фаєрволами [5]. Система відзначається високою точністю виявлення атак, низьким рівнем хибно-позитивних спрацьовувань. Та забезпечує ефективний захист серверів у реальному часі для забезпечення безперервності бізнес-процесів та запобігання фінансовим і репутаційним втратам.
- 31 December 2024
- Published by State University "Kyiv Aviation Institute" in Ukrainian Information Security Research Journal
- Vol. 26 (2) , 270-277
Open AccessСИСТЕМА РЕАЛІЗАЦІЇ ЗАХИСТУ СЕРВЕРІВ З УРАХУВАННЯМ АНОМАЛІЙ В ПАКЕТАХ
У сучасних умовах зростання кіберзагроз захист серверів стає критично важливим аспектом інформаційної безпеки, особливо в контексті збільшення обсягів мережевого трафіку та складності атак. Одним із ефективних підходів є використання систем захисту, що враховують аномалії у мережевих пакетах. Виявлення та обробка таких аномалій дозволяють оперативно ідентифікувати та нейтралізувати загрози, серед яких особливе місце займають DDoS-атаки. У статті розглянуто методи аналізу мережевого трафіку в реальному часі, що базуються на статистичних методах, а також алгоритмах машинного навчання для класифікації мережевих пакетів за їх поведінковими характеристиками [1]. Представлена система реалізує багаторівневий підхід до захисту серверів, який включає три основні етапи: первинну фільтрацію даних, статистичний аналіз та використання моделей машинного навчання. На першому етапі відсіюються шкідливі пакети на основі простих критеріїв, таких як заборонені IP-адреси або некоректний формат пакетів [2]. На другому етапі застосовується статистичний аналіз для виявлення відхилень у розподілі трафіку, наприклад, раптове збільшення кількості запитів або зміну розміру пакетів [3]. Третій етап передбачає застосування класифікаторів, які навчаються на історичних даних для визначення аномалій у поведінці мережі. Перелік представлених моделей дозволяє адаптуватися до нових типів атак шляхом автоматичного оновлення [4]. Наведемо переваги представленої системи: виявляє як традиційні DDoS-атаки (сканування портів, експлуатація вразливостей мережевих протоколів та спроби ін’єкцій SQL-запитів), так і інші види загроз. По-друге, її інтеграція з існуючими інструментами моніторингу та фаєрволами. Також забезпечує простоту впровадження без значного збільшення витрат завдяки інтеграції з існуючими інструментами моніторингу та фаєрволами [5]. Система відзначається високою точністю виявлення атак, низьким рівнем хибно-позитивних спрацьовувань. Та забезпечує ефективний захист серверів у реальному часі для забезпечення безперервності бізнес-процесів та запобігання фінансовим і репутаційним втратам.