Application of artificial intelligence for detecting information-technical impacts

Purpose of the work was to substantiate the effectiveness of applying artificial intelligence techniques (machine learning and deep learning) for the timely detection of destructive information-technical impacts on critical infrastructure objects. Methods . An analysis of scientific sources has been conducted scientific sources, including cybersecurity surveys and standards, and conducted an experiment on a public network attack dataset (UNSW-NB15) using machine learning (Random Forest) and a deep neural network. Evaluation was based on metrics such as accuracy, detection recall, F1-score, etc. Results . ML/DL methods show significantly higher attack detection accuracy compared to traditional signature-based tools: ~96% accuracy was achieved on the UNSW-NB15 dataset using a neural network, versus ~70% for the signature approach. We demonstrate that deep learning enables discovery of previously unknown attacks (including sophisticated multi-vector APTs) by recognizing hidden anomalies, and that ensemble and federated approaches improve detection reliability and speed. Conclusion . Integrating AI techniques into security monitoring systems considerably increases the protection efficiency of critical systems by proactively identifying cyberattacks with minimal false alarms. The experimental results confirm the practical applicability of the chosen methods for securing network infrastructure (energy, communications, industrial IoT). However, further work is needed to ensure robustness against adversarial attacks and to uphold AI reliability principles.