Detecting digital dependence: Inferring government use of hyperscale cloud from internet infrastructure data
Year: 2026
Type: article
Abstract: Understanding where government digital services are hosted is critical for assessing technological dependence and countries' progress toward digital sovereignty, yet hosting arrangements are rarely disclosed and difficult to observe from the outside. We ask whether the hosting providers of government digital services can be inferred from publicly observable Internet data. To this end, we develop and publish a unique ground truth dataset of the verified hosting arrangements of 269 public sector digital services, primarily obtained through Freedom of Information (FOI) requests in three countries. The data reveal that 77% of sampled UK digital public services are hosted on US hyperscalers, compared to 64% in Finland and 33% in the Philippines. Using these verified labels, we evaluate whether supervised machine-learning models can infer a service's hosting provider type from its NS, MX, ASN, and TLS records. Even with a limited training set, the best model distinguishes US hyperscaler from non-hyperscaler hosting with 83% accuracy and 0.83 macro-F1, a 13-percentage-point improvement over the single-feature heuristic inference used in prior large-scale studies. By demonstrating the feasibility of empirically estimating hosting dependencies from public data, the study opens a pathway toward systematic cross-national and longitudinal assessment of the extent of government dependence on foreign hosting infrastructures.
Source: Government Information Quarterly
Institutions: Aalto University, EADA Business School
Language: English
Field: Social Sciences
Domain: Social Sciences
Open Access status: hybrid
Estimated APC paid: $4,360
OpenAlex ID: W7237213124
