Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

From Privacy to Trust in the Agentic Era: A Taxonomy of Challenges in Trustworthy Federated Learning Through the Lens of Trust Report 2.0

Nuria Rodríguez-Barroso, Mario García-Márquez, M. Victoria Luzón, Francisco Herrera

TL;DR

This work proposes a requirement-driven taxonomy of challenges grounded in TAI and explicitly extended to account for control-plane decisions, agency, and system dynamics across the federated lifecycle, and introduces a coordination blueprint that structures cross-requirement trade-offs, decision justification, and governance alignment in TFL systems.

Abstract

Federated Learning (FL) enables privacy-preserving collaborative learning, yet deployments increasingly show that privacy guarantees alone do not sustain trust in high-risk settings. As FL systems move toward agentic AI, large language model-enabled, and dynamically adaptive architectures, trustworthiness becomes a system-level problem shaped by autonomous decision-making, non-stationary environments, and multi-stakeholder governance. We argue for Trustworthy FL (TFL), treating trust as a continuously maintained operating condition rather than a static model property. Through the lens of Trust Report 2.0, we propose a requirement-driven taxonomy of challenges grounded in TAI and explicitly extended to account for control-plane decisions, agency, and system dynamics across the federated lifecycle. Building on this diagnosis, we introduce a coordination blueprint that structures cross-requirement trade-offs, decision justification, and governance alignment in TFL systems. To operationalize assurance, Trust Report 2.0 is instantiated as a lightweight, privacy-preserving artifact that surfaces decision-centric trust evidence without centralizing raw data. We illustrate applicability via healthcare as a stress-test domain, focusing on oncology FL under regulatory pressure and clinical risk.

From Privacy to Trust in the Agentic Era: A Taxonomy of Challenges in Trustworthy Federated Learning Through the Lens of Trust Report 2.0

TL;DR

This work proposes a requirement-driven taxonomy of challenges grounded in TAI and explicitly extended to account for control-plane decisions, agency, and system dynamics across the federated lifecycle, and introduces a coordination blueprint that structures cross-requirement trade-offs, decision justification, and governance alignment in TFL systems.

Abstract

Federated Learning (FL) enables privacy-preserving collaborative learning, yet deployments increasingly show that privacy guarantees alone do not sustain trust in high-risk settings. As FL systems move toward agentic AI, large language model-enabled, and dynamically adaptive architectures, trustworthiness becomes a system-level problem shaped by autonomous decision-making, non-stationary environments, and multi-stakeholder governance. We argue for Trustworthy FL (TFL), treating trust as a continuously maintained operating condition rather than a static model property. Through the lens of Trust Report 2.0, we propose a requirement-driven taxonomy of challenges grounded in TAI and explicitly extended to account for control-plane decisions, agency, and system dynamics across the federated lifecycle. Building on this diagnosis, we introduce a coordination blueprint that structures cross-requirement trade-offs, decision justification, and governance alignment in TFL systems. To operationalize assurance, Trust Report 2.0 is instantiated as a lightweight, privacy-preserving artifact that surfaces decision-centric trust evidence without centralizing raw data. We illustrate applicability via healthcare as a stress-test domain, focusing on oncology FL under regulatory pressure and clinical risk.

Paper Structure

This paper contains 78 sections, 6 figures, 5 tables.

Table of Contents

  1. Introduction: From Privacy Guarantees to Trust Evidence
  2. Background on Federated Learning
  3. Why?
  4. How?
  5. A Requirement-Driven Taxonomy of Trust-Critical Challenges in Federated Learning
  6. Requirement 1: Human agency and oversight
  7. Challenge 1.1: Human-in-the-loop
  8. Challenge 1.2: Audience's understanding
  9. Challenge 1.3: Inference in Vertical FL
  10. Challenge 1.4: Enforcing Autonomy Boundaries and Oversight Gates
  11. Requirement 2: Technical robustness and safety
  12. Challenge 2.1: Poisoning attacks
  13. Challenge 2.2: Free riders
  14. Challenge 2.3: Out Of Distribution Detection
  15. Challenge 2.4: Evaluation Integrity under Agentic AI Coordination
  16. ...and 63 more sections

Figures (6)

  • Figure 1: Representation of the round in FL. Figure inspired by luzon2024tutorial.
  • Figure 2: Taxonomy of challenges to align FL with TAI. Inspired in diaz2023connecting, under each TAI requirement, we list the specific challenges involved to meet the requirement within an FL paradigm.
  • Figure 3: Visual representation of the main key challenges in TFL. For the sake of clarity, we use the same color theme that in Figure \ref{['fig:example']}.
  • Figure 4: Conceptual separation between the learning plane and the control plane in agentic FL systems. Feedback and signals exchanged between the two planes enable adaptive control, continuous monitoring, and the generation of auditable evidence that underpins trustworthiness over time.
  • Figure 5: Recurring coordination loop for TFL.
  • ...and 1 more figures