News-Based Cyber Attack Impact Assessment: Evaluating the Performance of Large Language Models

Growing cyber attack threat necessitates quick and reliable impact assessment methodologies. Traditional approaches, often dependent on expert analysis and established cybersecurity frameworks, are prone to human error, manual effort, and high costs. This study investigates the novel method of integration of Large Language Models (LLMs) with traditional cybersecurity frameworks to enhance the accuracy and efficiency of cyber attack impact assessments. This research explores LLMs' capabilities in processing unstructured text data from news articles to assess the impact of a cyber attack and evaluate various cost metrics. The performance of LLMs is evaluated through both quantitative analysis using the Mean Absolute Percentage Error (MAPE) and qualitative assessments via structured questionnaires comparing LLM outputs with expert evaluations. The findings indicate significant potential for this novel approach in impact assessment, though further research is necessary to prove its applicability.