ASSESSMENT OF THE EFFECTIVENESS OF SPECIALIZED ARTIFICIAL INTELLIGENCE MODELS IN THE CYBER DEFENSE SYSTEM OF UKRAINE: TECHNICAL AND LEGAL ASPECT
DOI:
https://doi.org/10.28925/2663-4023.2026.34.1237Keywords:
cybersecurity; artificial intelligence; GPT-5.4-Cyber; Anthropic Mythos; cyber threats; information security; law enforcement; National Police of Ukraine; EAI model; hybrid warfare.Abstract
The article explores the possibilities of using specialized artificial intelligence (AI) models in the field of cybersecurity, in particular, using the latest GPT-5.4-Cyber model as an example. Modern cyber threats are characterized by a high level of complexity, dynamism, and cross-border nature, which creates an urgent need for the use of intelligent systems for automated analysis and rapid response.
This issue becomes particularly relevant in the context of martial law and hybrid warfare, where Ukraine's digital infrastructure is becoming the object of constant attacks. The study analyzes in detail the functional features of the model, its role in ensuring national cyber defense, as well as potential risks and prospects for its implementation in the daily activities of law enforcement agencies of Ukraine. Special attention is paid to the activities of units of the National Police of Ukraine, which perform key functions in combating cybercrime, protecting state information resources, and maintaining law and order in the digital environment.
An author's mathematical model for a comprehensive assessment of the effectiveness of intelligent security systems (E AI) is proposed, which is based on the calculation of an integral indicator. This indicator combines such critical parameters as the accuracy of anomaly detection (Дetection), the speed of automated response (Response), the level of system autonomy (Automation), and resistance to external manipulation or data poisoning (Stability). The scientific and methodological approach to the use of neural network architecture by substantiating the feasibility of combined use of highly specialized and universal models. The results of the calculations and testing demonstrate a significant advantage of specialized models over general-purpose solutions due to their preliminary training on target data sets: international vulnerability registers (CVE) and current samples of malicious code.
The analysis of regulatory support shows that the implementation of such technologies should be strictly based on international standards ISO/IEC 27001:2022 and NIST CSF 2.0. It was determined that a high level of automation of decision-making requires a systematic update of the legislative framework of Ukraine to clearly define the legal status and responsibility for decisions made using artificial intelligence algorithms. Further scientific research should be aimed at developing and implementing closed departmental circuits for the deployment of such models without the risk of leakage of official and confidential information through public cloud services.
Downloads
References
Apruzzese, G., et al. (2023). The role of artificial intelligence in cybersecurity: A survey. ACM Computing Surveys, 55(4), Article 58. https://doi.org/10.1145/3546933
Buczak, A. L., & Guven, E. (2016). A survey of data mining and machine learning methods for cyber security intrusion detection. IEEE Communications Surveys & Tutorials, 18(2), 1153–1176. https://doi.org/10.1109/COMST.2015.2494502
Ferrag, M. A., et al. (2024). Revolutionizing cybersecurity with large language models: A comprehensive survey. IEEE Access, 12, 10415–10444. https://doi.org/10.1109/ACCESS.2024.3354356
International Organization for Standardization. (2022). Information security, cybersecurity and privacy protection — Information security management systems — Requirements (ISO/IEC Standard No. 27001:2022). https://www.iso.org/standard/27001
MITRE. (2024). MITRE ATT&CK® v15. https://attack.mitre.org/
National Institute of Standards and Technology. (2024). The NIST Cybersecurity Framework (CSF) 2.0. U.S. Department of Commerce. https://doi.org/10.6028/NIST.CSWP.29
NV Techno. (2024, May 14). GPT-5.4-Cyber: Shcho vmiie nova AI-model dlia kiberbezpeky [GPT-5.4-Cyber: What the new AI model for cybersecurity can do]. https://techno.nv.ua/ukr/it-industry/gpt-5-4-cyber-shcho-vmiye-nova-shi-model-vid-openai-dlya-kiberbezpeki-50600407.html
President of Ukraine. (2017). Pro Doktrynu informatsiinoi bezpeky Ukrainy [On the Doctrine of Information Security of Ukraine] (Decree No. 47/2017). https://zakon.rada.gov.ua/laws/show/47/2017
President of Ukraine. (2021). Pro rishennia Rady natsionalnoi bezpeky i oborony Ukrainy vid 14 travnia 2021 roku “Pro Stratehiiu kiberbezpeky Ukrainy” [On the decision of the National Security and Defense Council of Ukraine of May 14, 2021 “On the Cybersecurity Strategy of Ukraine”] (Decree No. 447/2021). https://zakon.rada.gov.ua/laws/show/447/2021
Sommer, R., & Paxson, V. (2010). Outside the closed world: On using machine learning for network intrusion detection. In 2010 IEEE Symposium on Security and Privacy (pp. 305–316). https://doi.org/10.1109/SP.2010.25
Verkhovna Rada of Ukraine. (1992). Pro informatsiiu [On information] (Law No. 2657-XII). https://zakon.rada.gov.ua/laws/show/2657-12
Verkhovna Rada of Ukraine. (1994). Pro zakhyst informatsii v informatsiino-komunikatsiinykh systemakh [On the protection of information in information and communication systems] (Law No. 80/94-VR). https://zakon.rada.gov.ua/laws/show/80/94-вр
Verkhovna Rada of Ukraine. (2017). Pro osnovni zasady zabezpechennia kiberbezpeky Ukrainy [On the basic principles of ensuring cybersecurity of Ukraine] (Law No. 2163-VIII). https://zakon.rada.gov.ua/laws/show/2163-19
Published
How to Cite
Issue
Section
License
Copyright (c) 2026 Світличний Віталій, Клімушин Петро, Онищенко Юрій

This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.