Sergiy Paliy
Research Subjects
Sergiy Paliy has a total of 21 co-authors. Here are the top co-authors by number of shared publications:
- Yaroslav Hozak
Київський національний університет імені Тараса Шевченка
3 shared publications - Myroslava Gladka
Taras Shevchenko National University of Kyiv; National Technical University of Ukraine “Igor Sikorsky Kyiv Polytechnic Institute
2 shared publications Сергій Палій
Київський національний університет ім. Тараса Шевченка, Київ
2 shared publicationsAlexander Kuchansky
Kyiv National University of Construction and Architecture Povitroflotskyi ave., 31, Kyiv, Ukraine, 03037
1 shared publication- Andrii Biloshchytskyi
University Administration, Astana IT University, Astana 010000, Kazakhstan
1 shared publication - Artem Biloshchytskyi
Astana IT University; Qazaq AI Research University
1 shared publication
Latest Publications
- 13 May 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-5
ROI-SSD: Region of Interest Single Shot Detector
- 13 May 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-5
ROI-SSD: Region of Interest Single Shot Detector
- Open Access
- 25 December 2025
- Published by Kyiv National University of Construction and Architecture in Management of Development of Complex Systems
- No. 64
Using the video channel of ATM devices to enhance the level of cybersecurity during customer interactions
The rapid growth of automated banking and cardless financial transactions has increased the need for advanced user authentication mechanisms that can withstand fraud, social engineering, and biometric spoofing. Traditional authentication methods based solely on cards and PINs are increasingly vulnerable to theft, duress, and replay attacks. In this context, biometric technologies, including facial recognition combined with real-time behavioral analysis, represent a promising direction for enhancing transaction security. This study proposes an integrated method to improve user authentication in automated financial terminals by utilizing the device’s video feed. The approach focuses on continuous identity verification and action monitoring during user interaction, supported by a formally modeled interaction protocol between the terminal and the decision-making system. The proposed method integrates videobased biometric identification, activity detection, and behavioral analysis into a single authentication workflow. A protocol model was developed that defines secure data exchange, time constraints, and decision logic between the user, the terminal, and the server component. The model supports adaptive decision outcomes, including approval, re-verification, or rejection, based on trust levels derived from biometric and behavioral features. The use of edge computing principles reduces latency and network load by transmitting compact feature vectors instead of raw video streams. The approach also includes degradation modes for network failures, anomaly detection mechanisms, and compliance-oriented data minimization strategies. A comparative analysis with existing global practices demonstrates that the proposed solution is in line with current trends in multifactor authentication, extending them through protocol formalization and real-time video analytics. The results show that integrated video-based authentication can significantly improve the security and resilience of financial terminals without compromising usability or accessibility. The proposed interoperability protocol enables scalable deployment, supports regulatory compliance, and provides the flexibility to adapt authentication levels to risk conditions. This method is applicable not only to ATMs but also to self-service kiosks, remote banking terminals, and mobile financial infrastructures, offering a robust foundation for next-generation secure financial transaction systems.
- 25 December 2025
- Published by Kyiv National University of Construction and Architecture in Management of Development of Complex Systems
- No. 64
Open AccessUsing the video channel of ATM devices to enhance the level of cybersecurity during customer interactions
The rapid growth of automated banking and cardless financial transactions has increased the need for advanced user authentication mechanisms that can withstand fraud, social engineering, and biometric spoofing. Traditional authentication methods based solely on cards and PINs are increasingly vulnerable to theft, duress, and replay attacks. In this context, biometric technologies, including facial recognition combined with real-time behavioral analysis, represent a promising direction for enhancing transaction security. This study proposes an integrated method to improve user authentication in automated financial terminals by utilizing the device’s video feed. The approach focuses on continuous identity verification and action monitoring during user interaction, supported by a formally modeled interaction protocol between the terminal and the decision-making system. The proposed method integrates videobased biometric identification, activity detection, and behavioral analysis into a single authentication workflow. A protocol model was developed that defines secure data exchange, time constraints, and decision logic between the user, the terminal, and the server component. The model supports adaptive decision outcomes, including approval, re-verification, or rejection, based on trust levels derived from biometric and behavioral features. The use of edge computing principles reduces latency and network load by transmitting compact feature vectors instead of raw video streams. The approach also includes degradation modes for network failures, anomaly detection mechanisms, and compliance-oriented data minimization strategies. A comparative analysis with existing global practices demonstrates that the proposed solution is in line with current trends in multifactor authentication, extending them through protocol formalization and real-time video analytics. The results show that integrated video-based authentication can significantly improve the security and resilience of financial terminals without compromising usability or accessibility. The proposed interoperability protocol enables scalable deployment, supports regulatory compliance, and provides the flexibility to adapt authentication levels to risk conditions. This method is applicable not only to ATMs but also to self-service kiosks, remote banking terminals, and mobile financial infrastructures, offering a robust foundation for next-generation secure financial transaction systems.
- Open Access
- 25 September 2025
- Published by Khmelnytskyi National University in Herald of Khmelnytskyi National University. Technical sciences
- Vol. 357 (5.1) , 86-98
ПОРІВНЯЛЬНИЙ АНАЛІЗ НЕЙРОННИХ МЕРЕЖ РОЗПІЗНАВАННЯ ОБ’ЄКТІВ НА ЗОБРАЖЕННЯХ
В роботі проведено аналіз ключових етапів розвитку нейронних мереж для розпізнавання об'єктів — від ранніх двоетапних архітектур (R-CNN, Fast R-CNN) до сучасних одноетапних моделей родини YOLO. Детально досліджено революційний перехід від пропозицій регіонів до end-to-end детекції, зокрема внесок YOLO в обробку зображень у реальному часі. Особливу увагу приділено еволюції архітектур YOLO: вдосконаленню точності, швидкодії та адаптації до обмежених ресурсів (від v1 до v8). Показано, як інновації — якірні рамки, FPN/PAN, CSPBackbone, квантування — вирішували проблеми попередників (недостатня швидкість, низька точність дрібних об'єктів, висока ресурсомісткість). Підкреслено практичну значущість сучасних YOLO-моделей у задачах автономного транспорту, систем безпеки та промислової автоматизації.
- 25 September 2025
- Published by Khmelnytskyi National University in Herald of Khmelnytskyi National University. Technical sciences
- Vol. 357 (5.1) , 86-98
Open AccessПОРІВНЯЛЬНИЙ АНАЛІЗ НЕЙРОННИХ МЕРЕЖ РОЗПІЗНАВАННЯ ОБ’ЄКТІВ НА ЗОБРАЖЕННЯХ
В роботі проведено аналіз ключових етапів розвитку нейронних мереж для розпізнавання об'єктів — від ранніх двоетапних архітектур (R-CNN, Fast R-CNN) до сучасних одноетапних моделей родини YOLO. Детально досліджено революційний перехід від пропозицій регіонів до end-to-end детекції, зокрема внесок YOLO в обробку зображень у реальному часі. Особливу увагу приділено еволюції архітектур YOLO: вдосконаленню точності, швидкодії та адаптації до обмежених ресурсів (від v1 до v8). Показано, як інновації — якірні рамки, FPN/PAN, CSPBackbone, квантування — вирішували проблеми попередників (недостатня швидкість, низька точність дрібних об'єктів, висока ресурсомісткість). Підкреслено практичну значущість сучасних YOLO-моделей у задачах автономного транспорту, систем безпеки та промислової автоматизації.
- Open Access
- 24 September 2025
- Published by Kyiv National University of Construction and Architecture in Management of Development of Complex Systems
- No. 63,p. 158-166
Optimizing healthcare delivery time on the battlefield using an IoT smart watch network and RNN LSTM
In military environments, emergencies, including combat injuries, are an integral part of combat operations. One of the main causes of soldier casualties on the front line is delayed medical assistance. Even when medical personnel arrive on time, they have a limited period known as the "golden hour" to provide quality and effective medical care. To enhance the efficiency of patient assessments in situations with limited time and resources, modern technologies and data analysis methods can be utilized. Specifically, smartwatches can be used to collect and transmit real-time data on a patient’s physiological parameters. These data can include heart rate, body temperature, blood oxygen level, blood pressure, and other vital indicators. The collected data can be analyzed using recurrent neural networks (RNNs), which are capable of identifying patterns and correlations between different physiological parameters and the patient's condition. Based on this analysis, a preliminary diagnosis can be established, aiding medical personnel in more quickly and accurately assessing the patient's condition and making appropriate decisions. Additionally, the medic will have the ability to view the patient’s health indicators in real-time. Before reaching the patient in the field, the medic will have an understanding of what has happened to the patient and know their current physiological indicators, thanks to the real-time data collection and analysis system. This approach helps reduce the time required to provide medical care while simultaneously increasing the likelihood of successful medical intervention. The obtained data can be used by medics to implement a medical triage system, allowing them to prioritize their efforts and resources. The aim of the article is to demonstrate an emergency management system in a military environment based on the Internet of Things (IoT) and machine learning methods—recurrent neural networks; to improve the efficiency of providing medical assistance to injured soldiers in the field; to reduce response time to emergencies; and to increase the survival chances of injured soldiers. Overall, the research highlights key aspects of optimizing medical assistance in situations with limited resources and time, and identifies ways for further development and implementation of technological innovations to enhance the effectiveness of this process.
- 24 September 2025
- Published by Kyiv National University of Construction and Architecture in Management of Development of Complex Systems
- No. 63,p. 158-166
Open AccessOptimizing healthcare delivery time on the battlefield using an IoT smart watch network and RNN LSTM
In military environments, emergencies, including combat injuries, are an integral part of combat operations. One of the main causes of soldier casualties on the front line is delayed medical assistance. Even when medical personnel arrive on time, they have a limited period known as the "golden hour" to provide quality and effective medical care. To enhance the efficiency of patient assessments in situations with limited time and resources, modern technologies and data analysis methods can be utilized. Specifically, smartwatches can be used to collect and transmit real-time data on a patient’s physiological parameters. These data can include heart rate, body temperature, blood oxygen level, blood pressure, and other vital indicators. The collected data can be analyzed using recurrent neural networks (RNNs), which are capable of identifying patterns and correlations between different physiological parameters and the patient's condition. Based on this analysis, a preliminary diagnosis can be established, aiding medical personnel in more quickly and accurately assessing the patient's condition and making appropriate decisions. Additionally, the medic will have the ability to view the patient’s health indicators in real-time. Before reaching the patient in the field, the medic will have an understanding of what has happened to the patient and know their current physiological indicators, thanks to the real-time data collection and analysis system. This approach helps reduce the time required to provide medical care while simultaneously increasing the likelihood of successful medical intervention. The obtained data can be used by medics to implement a medical triage system, allowing them to prioritize their efforts and resources. The aim of the article is to demonstrate an emergency management system in a military environment based on the Internet of Things (IoT) and machine learning methods—recurrent neural networks; to improve the efficiency of providing medical assistance to injured soldiers in the field; to reduce response time to emergencies; and to increase the survival chances of injured soldiers. Overall, the research highlights key aspects of optimizing medical assistance in situations with limited resources and time, and identifies ways for further development and implementation of technological innovations to enhance the effectiveness of this process.
- Open Access
- 27 June 2025
- Published by Kyiv National University of Construction and Architecture in Management of Development of Complex Systems
- No. 62,p. 125-133
Integrating IoT technologies for monitoring and managing urban civil defense shelters
In the current context of martial law and increased technological risks, the development of intelligent civil defense systems is becoming particularly relevant. This research is dedicated to the development of a comprehensive solution based on IoT technologies for monitoring and managing bomb shelters within urban infrastructure. An architecture for a distributed system is proposed, combining advanced sensor network technologies, reliable communication protocols, and intelligent data processing algorithms. The main scientific contribution of this work is the development of a holistic approach to creating a stable and effective monitoring infrastructure that considers the specific requirements for operation in extreme conditions. The system integrates a network of highly sensitive sensors to control microclimate parameters, including temperature, humidity, air composition, and radiation levels, as well as advanced methods for monitoring occupancy based on PIR sensors. An important aspect of the research is the development of a hybrid communication infrastructure that combines the advantages of different data transmission technologies to ensure system reliability in conditions of potential interference, including electronic warfare means. Special attention is paid to energy efficiency – the proposed adaptive power management algorithms allow for autonomous system operation over a prolonged period. The scientific novelty of the work lies in the development of a unified mathematical model for evaluating system reaction time, the creation of innovative load balancing algorithms under conditions of limited communication channel bandwidth, and the implementation of predictive analytics mechanisms for early detection of potential threats. The practical value of the research is manifested in its potential implementation by civil defense authorities, urban infrastructure developers, and security system manufacturers to create effective population protection systems in modern conditions.
- 27 June 2025
- Published by Kyiv National University of Construction and Architecture in Management of Development of Complex Systems
- No. 62,p. 125-133
Open AccessIntegrating IoT technologies for monitoring and managing urban civil defense shelters
In the current context of martial law and increased technological risks, the development of intelligent civil defense systems is becoming particularly relevant. This research is dedicated to the development of a comprehensive solution based on IoT technologies for monitoring and managing bomb shelters within urban infrastructure. An architecture for a distributed system is proposed, combining advanced sensor network technologies, reliable communication protocols, and intelligent data processing algorithms. The main scientific contribution of this work is the development of a holistic approach to creating a stable and effective monitoring infrastructure that considers the specific requirements for operation in extreme conditions. The system integrates a network of highly sensitive sensors to control microclimate parameters, including temperature, humidity, air composition, and radiation levels, as well as advanced methods for monitoring occupancy based on PIR sensors. An important aspect of the research is the development of a hybrid communication infrastructure that combines the advantages of different data transmission technologies to ensure system reliability in conditions of potential interference, including electronic warfare means. Special attention is paid to energy efficiency – the proposed adaptive power management algorithms allow for autonomous system operation over a prolonged period. The scientific novelty of the work lies in the development of a unified mathematical model for evaluating system reaction time, the creation of innovative load balancing algorithms under conditions of limited communication channel bandwidth, and the implementation of predictive analytics mechanisms for early detection of potential threats. The practical value of the research is manifested in its potential implementation by civil defense authorities, urban infrastructure developers, and security system manufacturers to create effective population protection systems in modern conditions.
- Open Access
- 29 November 2024
- Published by Kyiv National University of Construction and Architecture in Management of Development of Complex Systems
- No. 60,p. 221-229
Modern small networks for image classification. Feature analysis
With the increasing demand for deploying deep learning models on resource-constrained devices, such as smartphones, IoT sensors, and edge computing platforms, the need for efficient convolutional neural networks (CNNs) has become paramount. This paper offers a comprehensive review of several state-of-the-art lightweight CNN architectures designed to address these challenges by reducing computational complexity and memory usage, while maintaining competitive performance in image classification tasks. Key architectures reviewed include MobileNets, ShuffleNet, DiceNet, and ESPNet, each of which employs distinct strategies to optimize network efficiency. MobileNets introduce the concept of depthwise separable convolutions, which decompose the standard convolution operation into a depthwise convolution and a point-wise convolution (1x1). This drastically reduces the number of parameters and computations compared to traditional convolutions. ShuffleNet, on the other hand, leverages group convolutions and channel shuffling to enhance efficiency, allowing feature maps to be split and recombined, which reduces computational cost without significantly compromising accuracy. DiceNet builds upon these concepts by introducing multi-branch architecture with different dilation rates to capture features at multiple scales, enhancing both accuracy and efficiency in low-resource environments. ESPNet employs efficient spatial pyramidal structures, along with point-wise convolutions, to handle diverse spatial features at different scales while being highly computationally efficient. Despite these advancements, a common bottleneck across these architectures is the reliance on point-wise (1x1) convolutions, which, while more efficient than standard convolutions, still contribute significantly to the overall computational cost, particularly in deeper layers of the network. Furthermore, filter sizes are often optimized for performance in a cloud-based setting but may not be ideal for edge environments where computational and energy efficiency are crucial. We see the potential in changing filter sizes in some layers to 2x2 which is the smallest possible filter for spatial information extraction. Also it worth paying attention to the way the information is spread across channels as well as how the channels number is formed by replacing 1x1 convolution with a generic but yet predictable mathematical operation.
- 29 November 2024
- Published by Kyiv National University of Construction and Architecture in Management of Development of Complex Systems
- No. 60,p. 221-229
Open AccessModern small networks for image classification. Feature analysis
With the increasing demand for deploying deep learning models on resource-constrained devices, such as smartphones, IoT sensors, and edge computing platforms, the need for efficient convolutional neural networks (CNNs) has become paramount. This paper offers a comprehensive review of several state-of-the-art lightweight CNN architectures designed to address these challenges by reducing computational complexity and memory usage, while maintaining competitive performance in image classification tasks. Key architectures reviewed include MobileNets, ShuffleNet, DiceNet, and ESPNet, each of which employs distinct strategies to optimize network efficiency. MobileNets introduce the concept of depthwise separable convolutions, which decompose the standard convolution operation into a depthwise convolution and a point-wise convolution (1x1). This drastically reduces the number of parameters and computations compared to traditional convolutions. ShuffleNet, on the other hand, leverages group convolutions and channel shuffling to enhance efficiency, allowing feature maps to be split and recombined, which reduces computational cost without significantly compromising accuracy. DiceNet builds upon these concepts by introducing multi-branch architecture with different dilation rates to capture features at multiple scales, enhancing both accuracy and efficiency in low-resource environments. ESPNet employs efficient spatial pyramidal structures, along with point-wise convolutions, to handle diverse spatial features at different scales while being highly computationally efficient. Despite these advancements, a common bottleneck across these architectures is the reliance on point-wise (1x1) convolutions, which, while more efficient than standard convolutions, still contribute significantly to the overall computational cost, particularly in deeper layers of the network. Furthermore, filter sizes are often optimized for performance in a cloud-based setting but may not be ideal for edge environments where computational and energy efficiency are crucial. We see the potential in changing filter sizes in some layers to 2x2 which is the smallest possible filter for spatial information extraction. Also it worth paying attention to the way the information is spread across channels as well as how the channels number is formed by replacing 1x1 convolution with a generic but yet predictable mathematical operation.
- Open Access
- 13 December 2017
- Published by Private Company Technology Center in Eastern-European Journal of Enterprise Technologies
- Vol. 6 (2 (90)) , 16-22
A method to evaluate the scientific activity quality of HEIs based on a scientometric subjects presentation model
A method to evaluate the scientific activity quality of HEIs based on a scientometric subjects presentation model
- 13 December 2017
- Published by Private Company Technology Center in Eastern-European Journal of Enterprise Technologies
- Vol. 6 (2 (90)) , 16-22
Open AccessA method to evaluate the scientific activity quality of HEIs based on a scientometric subjects presentation model
A method to evaluate the scientific activity quality of HEIs based on a scientometric subjects presentation model