AI-генератор атак як інструмент проактивного кібер-тестування державних цифрових сервісів

Abstract

The paper substantiates the concept of an AI-based attack generator as a proactive cybersecurity testing instrument for governmental digital services. Traditional threat modeling and periodic penetration testing lack scalability and predictive capability in the context of rapidly evolving and AI-assisted cyber threats. The proposed approach employs large language models to generate structured and machine-readable attack scenarios based on the architectural description of public digital platforms. The generator enables dynamic combination of tactics and techniques aligned with MITRE ATT&CK and STRIDE models, facilitating multi-step attack path modeling. Unlike static security assessment methods, the AI-based mechanism supports continuous expansion of the attack surface analysis and risk-oriented prioritization of protection measures. The proposed concept contributes to the transition from reactive to adaptive and proactive cyber resilience in digital government infrastructures.