Abstract
The Internet's importance for promoting free and open communication has led to widespread crackdowns on its use in countries around the world. In this study, we investigate the relationship between national policies around freedom of speech and Internet topology in countries around the world. We combine techniques from network measurement and machine learning to identify features of Internet structure at the national level that are the best indicators of a country's level of freedom. We find that IP density and path lengths to other countries are the best indicators of a country's freedom. We also find that our methods predict freedom categories for countries with 91% accuracy.
Introduction
Measuring "Internet Freedom" has traditionally been a qualitative, labor-intensive process. Organizations like Freedom House rely on expert analysis of legal frameworks and reported censorship incidents. However, this approach often suffers from data gaps in regions where on-the-ground reporting is difficult. This research explores whether a nation's political stance on information control leaves a measurable footprint on its physical Internet infrastructure. By quantifying the relationship between network topology and political freedom, we can potentially use network structure as a proxy for freedom levels.
Methodology
To bridge the gap between "bits and ballots," we combined large-scale network measurement with machine learning.
Data Collection and Feature Engineering
We utilized the CAIDA Autonomous System (AS) relationship dataset as a baseline. To address known visibility issues, we employed the RIPE Atlas platform to perform thousands of targeted traceroutes, uncovering nearly 6,000 previously hidden network edges. From these augmented graphs, we extracted 20 features across four categories: structural metrics (e.g., transitivity), international connectivity (e.g., path lengths), IP demographics (IP density), and BGP routing properties.
Model Selection
We evaluated several machine learning models to predict Freedom of the Press Index (FPI) scores. The most effective model was a Decision Tree with Linear Regression at the leaves (DTLR). This approach groups countries with similar topological characteristics and applies a specific regression formula to each group.
Results
The DTLR model significantly outperformed standard linear regression and LASSO models. As shown in the prediction error distribution, the DTLR model exhibits a much tighter peak around zero error compared to simpler linear models.
Quantitatively, the DTLR model achieved:
- An average prediction error of only 7.06% for FPI scores.
- 81% accuracy in predicting specific freedom categories (Free, Partly-Free, Not Free).
- 91% accuracy when the task was simplified to distinguishing between "Free/Partly-Free" and "Not Free" regimes.
Key Insights and Discussion
Our analysis revealed several critical relationships between infrastructure and policy:
- IP Density as a Primary Indicator: IP density emerged as the strongest predictor. High IP-to-population ratios correlate strongly with freedom, likely reflecting both economic development and a lack of centralized IP management used for state control.
- Path Length and Control: We found a negative correlation between freedom and the maximum length of BGP-compliant paths to other countries. Longer paths often indicate centralized chokepoints, which are easier for governments to monitor or filter.
- Infrastructure of Censorship: Restrictive regimes typically exhibit low transitivity (fewer redundant ISP connections) and high centrality, facilitating country-wide shutdowns or deep-packet inspection.
- Outlier Analysis: The model identified interesting outliers like Singapore and Botswana. Singapore has a highly connected topology that suggests high freedom, yet it has a low FPI score due to restrictive media laws. Conversely, Botswana maintains high freedom despite a "restricted-looking" topology, likely due to infrastructure lagging behind political progress.
Conclusion
This study demonstrates that a country’s level of Internet freedom is physically reflected in its network topology. While the current model identifies correlations rather than direct causality, it provides a powerful new tool for monitoring global Internet freedom. Future work will focus on temporal analysis to track how network structures evolve in response to changing political climates.
Summary updated