Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Building Privacy-and-Security-Focused Federated Learning Infrastructure for Global Multi-Centre Healthcare Research

Fan Zhang, Daniel Kreuter, Javier Fernandez-Marques, BloodCounts Consortium, Gregory Verghese, Bernard Butler, Nicholas Lane, Suthesh Sivapalaratnam, Joseph Taylor, Norbert C. J. de Wit, Nicholas S. Gleadall, Carola-Bibiane Schönlieb, Michael Roberts

TL;DR

Results show that enforceable governance can function as a first-class privacy-preserving control, improving trustworthiness for scalable artificial intelligence (AI) in cross-jurisdictional healthcare deployments.

Abstract

Collaborative healthcare research across multiple institutions increasingly requires diverse clinical datasets, but cross-border data sharing is strictly constrained by privacy regulations. Federated learning (FL) enables model training while keeping data local; however, many existing frameworks remain proof-of-concept and do not adequately address governance risks such as unauthorised participation, misuse, and lack of accountability. In particular, enforceable mechanisms for authentication, authorisation, and accounting (AAA) are often missing, limiting real-world clinical deployment. This paper presents FLA$^3$ (Federated Learning with Authentication, Authorisation, and Accounting), a governance-aware federated learning platform that operationalises regulatory obligations through runtime policy enforcement. FLA$^3$ integrates eXtensible Access Control Markup Language (XACML) compliant attribute-based access control (ABAC), cryptographic accounting, and study-scoped federation directly into the federated learning orchestration layer to enforce institutional sovereignty and protocol adherence. We evaluate FLA$^3$ through two complementary studies. First, we demonstrate operational feasibility by deploying the platform infrastructure across five BloodCounts! Consortium institutions in four countries: United Kingdom, Netherlands, India, and The Gambia. Second, we assess clinical utility using simulated federation of full blood count (FBC) data from 54,446 samples from 35,315 subjects across 25 centres in the INTERVAL study. Results show that FLA$^3$ achieves predictive performance comparable to centralised training while strictly enforcing governance constraints. These results show that enforceable governance can function as a first-class privacy-preserving control, improving trustworthiness for scalable artificial intelligence (AI) in cross-jurisdictional healthcare deployments.

Building Privacy-and-Security-Focused Federated Learning Infrastructure for Global Multi-Centre Healthcare Research

TL;DR

Results show that enforceable governance can function as a first-class privacy-preserving control, improving trustworthiness for scalable artificial intelligence (AI) in cross-jurisdictional healthcare deployments.

Abstract

Collaborative healthcare research across multiple institutions increasingly requires diverse clinical datasets, but cross-border data sharing is strictly constrained by privacy regulations. Federated learning (FL) enables model training while keeping data local; however, many existing frameworks remain proof-of-concept and do not adequately address governance risks such as unauthorised participation, misuse, and lack of accountability. In particular, enforceable mechanisms for authentication, authorisation, and accounting (AAA) are often missing, limiting real-world clinical deployment. This paper presents FLA (Federated Learning with Authentication, Authorisation, and Accounting), a governance-aware federated learning platform that operationalises regulatory obligations through runtime policy enforcement. FLA integrates eXtensible Access Control Markup Language (XACML) compliant attribute-based access control (ABAC), cryptographic accounting, and study-scoped federation directly into the federated learning orchestration layer to enforce institutional sovereignty and protocol adherence. We evaluate FLA through two complementary studies. First, we demonstrate operational feasibility by deploying the platform infrastructure across five BloodCounts! Consortium institutions in four countries: United Kingdom, Netherlands, India, and The Gambia. Second, we assess clinical utility using simulated federation of full blood count (FBC) data from 54,446 samples from 35,315 subjects across 25 centres in the INTERVAL study. Results show that FLA achieves predictive performance comparable to centralised training while strictly enforcing governance constraints. These results show that enforceable governance can function as a first-class privacy-preserving control, improving trustworthiness for scalable artificial intelligence (AI) in cross-jurisdictional healthcare deployments.
Paper Structure (30 sections, 4 equations, 4 figures, 1 table)

This paper contains 30 sections, 4 equations, 4 figures, 1 table.

Table of Contents

  1. Introduction
  2. Related Work
  3. FL in Healthcare and Privacy Preservation
  4. Governance, Access Control, and Deployment in FL
  5. Privacy and Governance Requirements and Threat Model
  6. Healthcare Governance Requirements
  7. Derived Governance Requirements
  8. Threat Model
  9. System Architecture
  10. FL Runtime
  11. Multi-Study Federation
  12. AAA Framework
  13. Authentication
  14. Authorisation
  15. Request Model
  16. ...and 15 more sections

Figures (4)

  • Figure 1: FLA3 system architecture. FLA3 comprises three layers: central coordination via SuperLink (server), site-local SuperNodes (gateways), and ephemeral ClientApp processes (study-specific execution). The governance layer enforces constraints through -compliant policy evaluation and cryptographic audit logging.
  • Figure 2: Per-centre distributions for individual training and FedMAP. Each point corresponds to a donation centre. Boxplots summarise the distribution across centres with whiskers extending to 1.5 interquartile ranges. The red dashed line indicates centralised reference model performance. rocauc, rocauc; higher values indicate better predictive performance.
  • Figure 3: Per-centre performance gain from federation (FedMAP minus individual ) plotted against baseline individual performance. Centres with lower baseline performance benefit most from federation, indicating mitigation of inter-centre heterogeneity. rocauc, rocauc.
  • Figure 4: Regional median for individual training (left) and FedMAP (right). Centres were aggregated by region using the median to illustrate geographic consistency of performance. Identical colour scales are used across both maps.