Year: 2025
Type: article
Abstract: The increasing fusion of the global economy with digital infrastructure has elevated cyber threats from isolated technical issues to significant drivers of financial instability.Despite this, a measurement gap persists, as conventional financial risk models are ill-equipped to handle the high-dimensional and non-linear nature of Cyber Threat Intelligence (CTI).This study bridges this gap by developing and validating a predictive framework that translates global CTI into quantitative forecasts of systemic financial risk.Using a comprehensive dataset of over 77,000 daily cyber threat observations across 225 countries from 2015 to 2024, we forecast the U.S. St.Louis Fed Financial Stress Index (STLFSI).We conduct a comparative analysis of advanced deep learning architectures, including a Temporal Fusion Transformer (TFT), against canonical machine learning ensembles.Our results show that a gradient-boosted model (XGBoost) decisively outperforms other models, achieving an R² of 0.9883 and RMSE of 0.0321 on the hold-out test set.Employing Explainable AI (XAI) techniques, we deconstruct the model's predictions and find that its success stems from capturing the complex, non-linear interaction between cyber threat levels and the pre-existing state of financial market stress.This research provides robust empirical evidence of the cyber-financial nexus, offering a novel, data-driven methodology for asset managers, regulators, and security leaders to proactively quantify and manage a critical 21st-century risk.
DOI: 10.56831/psen-07-224
Language: English
Subfield: Finance
Domain: Social Sciences
Open Access status: gold
OpenAlex ID: W4414417786
FWCI:
Cites:
Cited by:
Related to: