Cybercrime in Numbers: Data-Driven Cybersecurity Evaluation

Global cybercrime is rapidly escalating, threatening both digital infrastructure and societal stability, and thus necessitates advanced data-driven countermeasures. Although prior studies have explored either cybercrime distribution or the effectiveness of cybersecurity policies, few have fused both dimensions into a unified analytical framework. This study is the first to propose a “Distribution-Policy-Correlation” triad approach, integrating clustering-based distribution modeling, policy breakpoint regression, and interpretable machine learning, surpassing earlier work that typically focuses on only one dimension. By introducing XGBoost and Shapley values, we provide an in-depth explanation of how demographic variables such as gross domestic product and Engel coefficient drive cybercrime, an approach rarely seen in current literature. We adopt DBSCAN to identify attacker-victim aggregation patterns across nations. Drawing on six countries (including US, Japan, and India), we employ a breakpoint regression model to evaluate the dynamic effects of cybersecurity policies before and after implementation-finding a short-term decrease in cybercrime but a potential weakening of policy impact over time. By combining XGBoost with SHAP, we reveal that higher gross domestic product and broader internet penetration strongly correlate with increased cybercrime, whereas a higher Engel coefficient demonstrates a marked negative correlation.