A Privacy-Preserving Federated Data Exchange Architecture for Trusted Cross-Border Digital Services in the East African Community

Abstract

Purpose: Cross-border digital services are growing rapidly in the East African Community (EAC), with higher demand for secure, trustworthy data exchange and safeguarding of institutions' data ownership and privacy. Given these issues, this paper presents a Privacy-Preserving Federated Data Exchange Architecture (PP-FDEA). Methodology: The architecture integrates federated learning, decentralized data control, authentication, authorization, consent management, encryption and differential privacy, as well as auditable governance. Country Health Indicators data is distributed across virtual EAC nodes representing Kenya, Uganda and Tanzania to evaluate the proposed architecture. The data used for training are stored locally and the number of samples is about 775–780 per node and the distribution does not change even by 5 samples. Findings: The Multi-Round FedAvg achieves an independent-test accuracy of about 0.82 and a DP-FedAvg with noise multiplier 0.3 achieves an independent-test accuracy of 0.80. Latency is still low, between 0.50 and 0.61ms, while all five governance and security scenarios get a score of 100%. Unique Contribution to Theory, Practice and Policy: Overall, the results show the potential of PP-FDEA for providing secure, privacy preserving, scalable and trusted EAC cross-border digital services.