ScholarWorks@숙명여자대학교

초록

This study analyzes the accident response performance by applying the Ai REsilience System (ARES), an artificial intelligence resilience system, in large-scale cyber disasters such as the National Information Resource Service (NIRS) data center fire. ARES integrates public sentiment data and technology logs into equivalent first-class inputs, and implements a cycle of cyclic resilience of "prediction→response→ recovery→ learning" through a hybrid AI engine that combines time series prediction (LSTM/Transformer), abnormality detection (Autoencoder/Isolation Forest), risk classification (GBM/MLP), and explainable AI (SHAP) and SOAR-based orchestration. This paper compares technical and social indicators such as average detection time (MTTD), average recovery time (MTTR), civil complaint growth rate, and public satisfaction (CSAT) by parallel simulation of traditional response model and ARES-based response model on the same accident timeline. This study empirically suggests that technological recovery performance and social trust recovery can be improved when the use of early warning of CX signals and automated orchestration are combined.

키워드