Застосування нечіткої логіки для оцінки ризиків у системі адаптивного аудиту кібербезпеки

Abstract

The theses substantiate the integration of fuzzy logic mathematical tools into risk assessment procedures within adaptive cybersecurity audit systems. It is proven that shifting from binary models to fuzzy ones allows effective handling of uncertainty amidst heterogeneous threats. An improved model is proposed that transforms input parameters (threat level, asset criticality, vulnerabilities) into linguistic variables to calculate a dynamic integral risk indicator. Based on an analysis of recent studies (2023–2025), it is demonstrated that using knowledge bases with production rules and neuro-fuzzy networks (ANFIS) ensures anomaly detection accuracy up to 99.2% and significantly reduces false positives. The results confirm the efficacy of implementing this approach within ISO/IEC 27001 and NIST CSF frameworks to enhance critical infrastructure resilience.