Volodymyr Vasylenko
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Volodymyr Vasylenko has a total of 7 co-authors. Here are the top co-authors by number of shared publications:
- Ganna Grynkevych
Borys Grinchenko Kyiv metropolitan University
7 shared publications - Dmytro Rudin
Capital One (Plano, Texas, USA)
4 shared publications - Illia Kuznietsov
State University "Kyiv Aviation Institute
1 shared publication - Oleksandr Pliekhov
EPAM Systems (United States)
1 shared publication - Valeriy Savenok
State University of Information and Communication Technologies, Kyiv, Ukraine
1 shared publication - Yuliia Berezovska
State University of Information and Communication Technologies, Kyiv, Ukraine
1 shared publication
Latest Publications
- Open Access
- 26 October 2026
- Published by Borys Grinchenko Kyiv University in Cybersecurity: Education, Science, Technique
- Vol. 3 (35) , 688-702
A USAGE-AWARE METHOD FOR REPAIRING PRIVILEGE ESCALATION PATHS IN AWS IDENTITY AND ACCESS MANAGEMENT
Misconfigured identity policies in Amazon Web Services (AWS) let a low-privileged user or role obtain administrator rights through chains of individually harmless permissions such as iam:PassRole or iam:CreateAccessKey. Existing repair methods choose the fix by the number of policy changes, so a “minimal” repair can break deployments, service roles and people who use the permission for legitimate work. This paper proposes a usage-aware method for repairing privilege escalation paths in AWS Identity and Access Management (IAM). The account is modeled as a permission graph with monotone escalation rules, and usage evidence from access logs defines operational impact as the weighted number of observed (principal, action, resource) uses that a repair would break. Repair is solved as an exact counterexample-guided CP-SAT optimization that either minimizes impact under full safety or maximizes removed risk under a change budget; under stated monotonicity assumptions the result is optimal within the repair vocabulary. On 50 synthetic organizations the method removed all escalations while breaking 78% fewer legitimate uses than a minimum-size repair (32 versus 144; Wilcoxon test, p < 10⁻⁶), at the cost of a median of 6.5 instead of 3 edited policy statements. Hiding iam:PassRole from the evidence, as AWS last-accessed data does, raised impact from 1.21 to 2.18, and a conservative prior for unobservable actions brought it back to 1.21. With four changes the budgeted variant removed 96.7% of escalation risk. On a public export of the IAM Vulnerable account the model detected 50 of 58 in-scope escalation scenarios, and usage-aware repair reduced impact by 65.7% relative to a minimum-size repair.
- 26 October 2026
- Published by Borys Grinchenko Kyiv University in Cybersecurity: Education, Science, Technique
- Vol. 3 (35) , 688-702
Open AccessA USAGE-AWARE METHOD FOR REPAIRING PRIVILEGE ESCALATION PATHS IN AWS IDENTITY AND ACCESS MANAGEMENT
Misconfigured identity policies in Amazon Web Services (AWS) let a low-privileged user or role obtain administrator rights through chains of individually harmless permissions such as iam:PassRole or iam:CreateAccessKey. Existing repair methods choose the fix by the number of policy changes, so a “minimal” repair can break deployments, service roles and people who use the permission for legitimate work. This paper proposes a usage-aware method for repairing privilege escalation paths in AWS Identity and Access Management (IAM). The account is modeled as a permission graph with monotone escalation rules, and usage evidence from access logs defines operational impact as the weighted number of observed (principal, action, resource) uses that a repair would break. Repair is solved as an exact counterexample-guided CP-SAT optimization that either minimizes impact under full safety or maximizes removed risk under a change budget; under stated monotonicity assumptions the result is optimal within the repair vocabulary. On 50 synthetic organizations the method removed all escalations while breaking 78% fewer legitimate uses than a minimum-size repair (32 versus 144; Wilcoxon test, p < 10⁻⁶), at the cost of a median of 6.5 instead of 3 edited policy statements. Hiding iam:PassRole from the evidence, as AWS last-accessed data does, raised impact from 1.21 to 2.18, and a conservative prior for unobservable actions brought it back to 1.21. With four changes the budgeted variant removed 96.7% of escalation risk. On a public export of the IAM Vulnerable account the model detected 50 of 58 in-scope escalation scenarios, and usage-aware repair reduced impact by 65.7% relative to a minimum-size repair.
- Open Access
- 24 September 2026
- Published by Borys Grinchenko Kyiv University in Cybersecurity: Education, Science, Technique
- Vol. 2 (34) , 726-743
AN EXTENDED STRIDE THREAT MODEL FOR OFFLINE BLE MESH DISASTER COMMUNICATION APPLICATIONS
Offline mesh messaging over Bluetooth Low Energy (BLE) is increasingly used as a communication layer of last resort when infrastructure fails. Its security is normally analysed at the BLE stack level, and rarely at the level of a deployed application in a disaster context. This paper presents a systematic threat model for that system class. We extend STRIDE with two categories the disaster context requires and standard STRIDE cannot express - Bystander harm and Physical compromise - and apply a reduced three-factor rating scheme (Damage x Reproducibility x Affected parties) with published anchors. The methodology is instantiated on a single case study - ResQMesh AI, an open-source Android BLE mesh emergency-communication platform - which fixes the scope of the mitigation mapping. The catalogue contains 29 threats and six adversary profiles, with a per-threat mitigation mapping in which 3 threats are effectively addressed, 19 partially and 7 not at all. The two coupling effects the model predicts are checked by discrete-time simulation of a 60-node flooding mesh (TTL 7, 20 runs): forged critical-priority traffic at three times per-node transmission capacity reduces the deadline delivery ratio of genuine critical messages from 0.87 to 0.59, a per-identity relay budget restores it to 0.85, and Sybil identity rotation defeats the budget entirely; driving 40% of nodes into the lowest energy mode reduces delivery from 0.88 to 0.50, with the critical relay floor recovering 3-5 percentage points. Three findings generalise beyond the case study: removing infrastructure relocates the entire trust bootstrap onto a trust-on-first-use radio exchange, making spoofing the highest-ranked class; AI-driven message prioritisation creates an amplification primitive, since forged high-priority traffic pre-empts genuine emergency traffic; and energy-aware relaying gives an adversary a low-cost lever for partitioning the mesh.
- 24 September 2026
- Published by Borys Grinchenko Kyiv University in Cybersecurity: Education, Science, Technique
- Vol. 2 (34) , 726-743
Open AccessAN EXTENDED STRIDE THREAT MODEL FOR OFFLINE BLE MESH DISASTER COMMUNICATION APPLICATIONS
Offline mesh messaging over Bluetooth Low Energy (BLE) is increasingly used as a communication layer of last resort when infrastructure fails. Its security is normally analysed at the BLE stack level, and rarely at the level of a deployed application in a disaster context. This paper presents a systematic threat model for that system class. We extend STRIDE with two categories the disaster context requires and standard STRIDE cannot express - Bystander harm and Physical compromise - and apply a reduced three-factor rating scheme (Damage x Reproducibility x Affected parties) with published anchors. The methodology is instantiated on a single case study - ResQMesh AI, an open-source Android BLE mesh emergency-communication platform - which fixes the scope of the mitigation mapping. The catalogue contains 29 threats and six adversary profiles, with a per-threat mitigation mapping in which 3 threats are effectively addressed, 19 partially and 7 not at all. The two coupling effects the model predicts are checked by discrete-time simulation of a 60-node flooding mesh (TTL 7, 20 runs): forged critical-priority traffic at three times per-node transmission capacity reduces the deadline delivery ratio of genuine critical messages from 0.87 to 0.59, a per-identity relay budget restores it to 0.85, and Sybil identity rotation defeats the budget entirely; driving 40% of nodes into the lowest energy mode reduces delivery from 0.88 to 0.50, with the critical relay floor recovering 3-5 percentage points. Three findings generalise beyond the case study: removing infrastructure relocates the entire trust bootstrap onto a trust-on-first-use radio exchange, making spoofing the highest-ranked class; AI-driven message prioritisation creates an amplification primitive, since forged high-priority traffic pre-empts genuine emergency traffic; and energy-aware relaying gives an adversary a low-cost lever for partitioning the mesh.
- Open Access
- 25 June 2026
- Published by Borys Grinchenko Kyiv University in Cybersecurity: Education, Science, Technique
- Vol. 1 (33) , 234
LOCAL OPEN-SOURCE PLATFORM FOR CYBER INCIDENT MONITORING AND RESPONSE WITH AI SUPPORT AS AN ALTERNATIVE TO EDR/XDR SOLUTIONS
The paper presents an approach to designing a local cybersecurity monitoring and incident response platform based on open-source technologies with integrated artificial intelligence support. The relevance of the study is driven by the growing complexity of cyber threats, the increasing number of endpoints in modern infrastructures, and the need for continuous monitoring and rapid response within Security Operations Centers. Traditional EDR and XDR solutions provide high levels of automation and detection capabilities but rely on subscription-based models that introduce long-term operational constraints and limit flexibility in infrastructure management. The study examines the functional roles of EDR, XDR, SIEM, and SOC within enterprise environments and identifies key challenges related to scalability, data ownership, and dependency on proprietary platforms. Particular attention is given to the limitations of centralized commercial solutions when applied to distributed infrastructures with hundreds of managed endpoints. The paper proposes a model of a local cybersecurity platform that integrates endpoint monitoring, event correlation, and log management within a unified architecture. The proposed solution is based on Wazuh for endpoint detection and SIEM functions, OpenSearch-compatible indexing for event storage and analytics, and on-premises storage systems for secure and scalable log retention. A dedicated local AI analysis layer is introduced to support SOC processes by automating alert summarization, correlating related events, generating hypotheses about incident origins, and assisting analysts in decision-making without exposing sensitive data to external services. The results include a description of the system architecture, hardware configuration, data processing workflows, and storage strategies. The proposed approach emphasizes transparency, flexibility, and adaptability to specific organizational requirements. Additionally, the model incorporates a structured proof-of-concept methodology to evaluate detection coverage, system performance, and the effectiveness of AI-assisted workflows. The study demonstrates that the integration of open-source tools with local AI capabilities can enhance the efficiency of cybersecurity operations while maintaining control over data and infrastructure. The proposed platform provides a viable alternative to traditional EDR/XDR solutions in scenarios where customization, cost predictability, and data sovereignty are critical factors.
- 25 June 2026
- Published by Borys Grinchenko Kyiv University in Cybersecurity: Education, Science, Technique
- Vol. 1 (33) , 234
Open AccessLOCAL OPEN-SOURCE PLATFORM FOR CYBER INCIDENT MONITORING AND RESPONSE WITH AI SUPPORT AS AN ALTERNATIVE TO EDR/XDR SOLUTIONS
The paper presents an approach to designing a local cybersecurity monitoring and incident response platform based on open-source technologies with integrated artificial intelligence support. The relevance of the study is driven by the growing complexity of cyber threats, the increasing number of endpoints in modern infrastructures, and the need for continuous monitoring and rapid response within Security Operations Centers. Traditional EDR and XDR solutions provide high levels of automation and detection capabilities but rely on subscription-based models that introduce long-term operational constraints and limit flexibility in infrastructure management. The study examines the functional roles of EDR, XDR, SIEM, and SOC within enterprise environments and identifies key challenges related to scalability, data ownership, and dependency on proprietary platforms. Particular attention is given to the limitations of centralized commercial solutions when applied to distributed infrastructures with hundreds of managed endpoints. The paper proposes a model of a local cybersecurity platform that integrates endpoint monitoring, event correlation, and log management within a unified architecture. The proposed solution is based on Wazuh for endpoint detection and SIEM functions, OpenSearch-compatible indexing for event storage and analytics, and on-premises storage systems for secure and scalable log retention. A dedicated local AI analysis layer is introduced to support SOC processes by automating alert summarization, correlating related events, generating hypotheses about incident origins, and assisting analysts in decision-making without exposing sensitive data to external services. The results include a description of the system architecture, hardware configuration, data processing workflows, and storage strategies. The proposed approach emphasizes transparency, flexibility, and adaptability to specific organizational requirements. Additionally, the model incorporates a structured proof-of-concept methodology to evaluate detection coverage, system performance, and the effectiveness of AI-assisted workflows. The study demonstrates that the integration of open-source tools with local AI capabilities can enhance the efficiency of cybersecurity operations while maintaining control over data and infrastructure. The proposed platform provides a viable alternative to traditional EDR/XDR solutions in scenarios where customization, cost predictability, and data sovereignty are critical factors.
- Open Access
- 18 June 2026
- Published by Taras Shevchenko National University of Kyiv in Collection of scientific works of the Military Institute of Kyiv National Taras Shevchenko University
- No. 91,p. 24-32
CONCEPTUAL MODEL FOR THE APPLICATION OF ARTIFICIAL GENERAL INTELLIGENCE TOOLS IN SEARCH ENGINE OPTIMIZATION INTRODUCTION
Background. This article investigates the prospects of applying Artificial General Intelligence (AGI) to automate SEO analysis and content generation. The relevance of the study is determined by the need to improve the effectiveness of search engine optimization amid rapidly changing search engine algorithms, the growing volume of digital content, and the need to account for user behavior. The proposed conceptual model of SEO/AGI analysis involves the integration of Natural Language Processing (NLP), Machine Learning (ML), and Large Language Models (LLMs) into a unified adaptive system. Accordingly, the aim of the article is to develop a conceptual model of SEO/AGI analysis that formalizes the assessment of the semantic relevance of content, its ranking, generation, and optimization in accordance with user search intent. Methods. To develop the conceptual model of SEO/AGI analysis, methods of mathematical modeling, formalization, and optimization were employed. SEO/AGI analysis is considered as a problem of maximizing the Click-Through Rate (CTR) of search results. The model incorporates text vectorization and semantic relevance assessment using NLP, feature-space construction and a machine-learning-based ranking function, as well as conditional content generation and quality assessment using LLMs. To ensure the adaptability of the model, a feedback-loop mechanism is incorporated, enabling the system parameters to be adjusted based on user behavioral factors. Results. As a result of the study, a conceptual model of SEO/AGI analysis was developed that integrates NLP, ML, and LLM technologies into a unified sequential optimization process. The model encompasses text vectorization, semantic relevance assessment, feature-space construction, machine-based ranking, and content generation and quality assessment. The criteria for evaluating the integration of AGI components include cosine similarity, embeddings, and an SEO score, with the SEO score considered as an indicator directly characterizing the impact of AGI components on search visibility. The proposed adaptive learning mechanism enables iterative adjustment of the model parameters according to the results of user interactions with search results. Conclusions. The proposed conceptual model of SEO/AGI analysis formalizes the integration of NLP, ML, and LLM technologies into a unified adaptive system and provides a theoretical foundation for AI-oriented search engine optimization. The model combines semantic relevance assessment, ranking, content generation and quality evaluation, as well as the use of feedback based on user behavioral factors. The proposed approach provides a basis for transitioning from conventional ranking approaches toward user-centric SEO and may be applied in the further development of self-learning and adaptive SEO systems. A promising direction for future research is the validation of the proposed model through Python-based simulation using appropriate mathematical libraries.
- 18 June 2026
- Published by Taras Shevchenko National University of Kyiv in Collection of scientific works of the Military Institute of Kyiv National Taras Shevchenko University
- No. 91,p. 24-32
Open AccessCONCEPTUAL MODEL FOR THE APPLICATION OF ARTIFICIAL GENERAL INTELLIGENCE TOOLS IN SEARCH ENGINE OPTIMIZATION INTRODUCTION
Background. This article investigates the prospects of applying Artificial General Intelligence (AGI) to automate SEO analysis and content generation. The relevance of the study is determined by the need to improve the effectiveness of search engine optimization amid rapidly changing search engine algorithms, the growing volume of digital content, and the need to account for user behavior. The proposed conceptual model of SEO/AGI analysis involves the integration of Natural Language Processing (NLP), Machine Learning (ML), and Large Language Models (LLMs) into a unified adaptive system. Accordingly, the aim of the article is to develop a conceptual model of SEO/AGI analysis that formalizes the assessment of the semantic relevance of content, its ranking, generation, and optimization in accordance with user search intent. Methods. To develop the conceptual model of SEO/AGI analysis, methods of mathematical modeling, formalization, and optimization were employed. SEO/AGI analysis is considered as a problem of maximizing the Click-Through Rate (CTR) of search results. The model incorporates text vectorization and semantic relevance assessment using NLP, feature-space construction and a machine-learning-based ranking function, as well as conditional content generation and quality assessment using LLMs. To ensure the adaptability of the model, a feedback-loop mechanism is incorporated, enabling the system parameters to be adjusted based on user behavioral factors. Results. As a result of the study, a conceptual model of SEO/AGI analysis was developed that integrates NLP, ML, and LLM technologies into a unified sequential optimization process. The model encompasses text vectorization, semantic relevance assessment, feature-space construction, machine-based ranking, and content generation and quality assessment. The criteria for evaluating the integration of AGI components include cosine similarity, embeddings, and an SEO score, with the SEO score considered as an indicator directly characterizing the impact of AGI components on search visibility. The proposed adaptive learning mechanism enables iterative adjustment of the model parameters according to the results of user interactions with search results. Conclusions. The proposed conceptual model of SEO/AGI analysis formalizes the integration of NLP, ML, and LLM technologies into a unified adaptive system and provides a theoretical foundation for AI-oriented search engine optimization. The model combines semantic relevance assessment, ranking, content generation and quality evaluation, as well as the use of feedback based on user behavioral factors. The proposed approach provides a basis for transitioning from conventional ranking approaches toward user-centric SEO and may be applied in the further development of self-learning and adaptive SEO systems. A promising direction for future research is the validation of the proposed model through Python-based simulation using appropriate mathematical libraries.
- Open Access
- 26 June 2025
- Published by Borys Grinchenko Kyiv University in Cybersecurity: Education, Science, Technique
- Vol. 4 (28) , 137-149
The ROLE OF CAPTURE-THE-FLAG (CTF) CHALLENGES IN CYBERSECURITY RESEARCH AND TRAINING: ANALYSIS OF THE "EDITORIAL" MACHINE
Cybersecurity threats continue to evolve, requiring continuous improvements in training methodologies. Traditional theoretical education in cybersecurity often lacks practical engagement, leading to a skills gap in real-world security operations. Capture-the-Flag (CTF) challenges have emerged as an effective method for developing critical cybersecurity skills, offering participants a hands-on approach to penetration testing, network security, and privilege escalation techniques. This study explores the educational value of CTF challenges by analyzing the "Editorial" machine from the Hack The Box platform. The article provides a structured walkthrough, detailing key phases such as reconnaissance, exploitation, and privilege escalation. The exploitation phase demonstrates the identification of SQL injection vulnerabilities, while the privilege escalation phase highlights the risks of misconfigured Git repositories and sudo permissions. A scientific analysis of these vulnerabilities is presented, emphasizing their implications for real-world cybersecurity threats. The study also includes defensive strategies to mitigate such risks, advocating for secure coding practices, privilege management, and automated security audits. Additionally, the integration of CTF challenges into professional cybersecurity training is discussed, reinforcing their effectiveness in improving problem-solving skills and real-world preparedness. The findings support the growing role of CTFs in cybersecurity education and professional development. Future research may focus on enhancing the realism of CTF exercises, improving automation for skill assessment, and integrating these challenges into structured cybersecurity curricula. By bridging the gap between theoretical knowledge and hands-on experience, CTFs continue to play a vital role in developing a skilled cybersecurity workforce.
- 26 June 2025
- Published by Borys Grinchenko Kyiv University in Cybersecurity: Education, Science, Technique
- Vol. 4 (28) , 137-149
Open AccessThe ROLE OF CAPTURE-THE-FLAG (CTF) CHALLENGES IN CYBERSECURITY RESEARCH AND TRAINING: ANALYSIS OF THE "EDITORIAL" MACHINE
Cybersecurity threats continue to evolve, requiring continuous improvements in training methodologies. Traditional theoretical education in cybersecurity often lacks practical engagement, leading to a skills gap in real-world security operations. Capture-the-Flag (CTF) challenges have emerged as an effective method for developing critical cybersecurity skills, offering participants a hands-on approach to penetration testing, network security, and privilege escalation techniques. This study explores the educational value of CTF challenges by analyzing the "Editorial" machine from the Hack The Box platform. The article provides a structured walkthrough, detailing key phases such as reconnaissance, exploitation, and privilege escalation. The exploitation phase demonstrates the identification of SQL injection vulnerabilities, while the privilege escalation phase highlights the risks of misconfigured Git repositories and sudo permissions. A scientific analysis of these vulnerabilities is presented, emphasizing their implications for real-world cybersecurity threats. The study also includes defensive strategies to mitigate such risks, advocating for secure coding practices, privilege management, and automated security audits. Additionally, the integration of CTF challenges into professional cybersecurity training is discussed, reinforcing their effectiveness in improving problem-solving skills and real-world preparedness. The findings support the growing role of CTFs in cybersecurity education and professional development. Future research may focus on enhancing the realism of CTF exercises, improving automation for skill assessment, and integrating these challenges into structured cybersecurity curricula. By bridging the gap between theoretical knowledge and hands-on experience, CTFs continue to play a vital role in developing a skilled cybersecurity workforce.