Integration of Artificial Intelligence into Early Detection Systems for Cyber Threats for Critical Infrastructure in Wartime Conditions

Abstract

The relevance of the study is conditioned by the growing number of targeted cyber-attacks on energy, transport, and communications facilities in the context of hybrid threats. The purpose of the study was to develop and theoretically analyse the potential of an adaptive multi-level system for early detection of cyber threats that can function in conditions of limited resources and technical instability. As part of the study, a system architecture was built that combines network traffic analysis, evaluation of user behavioural patterns, fuzzy set logic for risk assessment, and self-learning modules that allow automatic updating of response rules. The effectiveness of the model was evaluated by comparative analysis of its characteristics with the data given in the scientific literature. The results of the study showed that the integration of convolutional neural networks (CNN), recurrent neural networks (RNN), and fuzzy logic into a multi-level early detection system for cyber threats allows effectively analysing network traffic and user behaviour. The proposed model demonstrated the ability to adapt to different types of threats, reducing the load on resources and ensuring high efficiency in an unstable infrastructure. The results of the study confirmed the prospects for integrating edge and fuzzy components into the cyber defence model. This approach increases the system's ability to detect complex threats in a timely manner, maintain its performance in crisis incidents, and develop adaptability to new types of attacks. Theoretical conclusions confirm the feasibility of implementing the proposed model for protecting critical infrastructure objects in the context of contemporary hybrid challenges.