Large Language Models (LLM) for Estimating the Cost of Cyber-attacks
A framework leveraging Large Language Models and big data analytics to estimate the financial impact of cyber threats, specifically focusing on lost business opportunities in the banking sector is presented, highlighting the superior accuracy of LLMs in estimating business activity disruptions.
Abstract
With the expansion of digital services and intelligent agents, cyber-attacks are increasingly frequent and impactful. Estimating the financial consequences of these attacks has become crucial in guiding investments in mitigation and defense strategies. This paper presents a framework leveraging Large Language Models (LLMs) and big data analytics to estimate the financial impact of cyber threats, specifically focusing on lost business opportunities in the banking sector. As a frequent target of cyberattacks, the banking industry suffers significant financial losses and a decline in customer trust. By analyzing over 23 billion transactions, the LLM algorithm identifies business activity patterns and calculates losses during operational downtimes. The study compares the performance of LLMs with alternative models, including Deep Learning, Support Vector Machines (SVM), and Random Walk, highlighting the superior accuracy of LLMs in estimating business activity disruptions. The findings provide a scalable methodology for calculating the financial cost of cyber-attacks in the banking sector, with potential applications in other industries. The study underscores the critical need for robust cybersecurity measures and effective risk mitigation strategies, given the high costs incurred during cyber-attack downtimes.