Prediction of Cybercrime and Policy Evaluation Using a Combination of PageRank and ARIMA-DID Algorithms

Abstract

This paper focuses on the distribution characteristics of global cybercrime and the effectiveness of policy interventions, constructing a multi-model analysis framework that integrates PageRank and ARIMA-DID. Firstly, the study uses the PageRank algorithm to calculate the cybercrime connectivity of countries, identifying the United States as the core risk node. Based on indicators such as cybersecurity preparedness and the number of cases, countries are categorized into high, medium, and low-risk groups. It is found that high-risk countries exhibit the characteristics of ‘high attack volume, high losses, and low cybersecurity preparedness.’ Further combining the ARIMA model with the difference-in-differences (DID) method, using the United States' 2013 cybersecurity policy as an example, the study constructs a counterfactual prediction scenario, confirming that policy implementation significantly reduces actual attack volumes compared to predicted values, but long-term effects are influenced by external factors. Cross-national comparisons show that low-risk countries like Japan and Australia have better policy sustainability, and preventive measures (like vulnerability management) have significantly higher cost-effectiveness than post-incident responses. The study provides a risk assessment model and policy effectiveness quantification tools for global cybersecurity governance.