Federated Unlearning in Edge Networks: A Survey of Fundamentals, Challenges, Practical Applications and Future Directions

TL;DR

Federated Unlearning (FUL) addresses Right-to-be-forgotten needs in edge Federated Learning by enabling selective removal of contributions from data samples, clients, or classes without full retraining. The standard FL update, often implemented as FedAvg, is $ \boldsymbol{\omega}_G^{t+1} = \sum_{n=1}^{N} p_n \boldsymbol{\omega}_n^t$ with $p_n = s_n / \sum_{n} s_n$. The survey organizes FUL methods into four families—provable server-side exactness, selective rollback strategies, clustering/partitioning with controlled participation, and compression/distillation/acceleration—and analyzes their edge-compatibility and trade-offs. It also discusses security/privacy considerations (poisoning defenses, verifiable deletion, privacy preservation, cryptographic protocols) and identifies open challenges such as stochasticity, non-IID data, and governance, proposing end-to-end erase pipelines as a research direction. The paper highlights real-world applications in cyberattack recovery, consent-aware caching/offloading, personalized on-device AI, and vehicular networks, underscoring the practical impact of regulation-compliant federated systems.

Abstract

The proliferation of connected devices and privacy-sensitive applications has accelerated the adoption of Federated Learning (FL), a decentralized paradigm that enables collaborative model training without sharing raw data. While FL addresses data locality and privacy concerns, it does not inherently support data deletion requests that are increasingly mandated by regulations such as the Right to be Forgotten (RTBF). In centralized learning, this challenge has been studied under the concept of Machine Unlearning (MU), that focuses on efficiently removing the influence of specific data samples or clients from trained models. Extending this notion to federated settings has given rise to Federated Unlearning (FUL), a new research area concerned with eliminating the contributions of individual clients or data subsets from the global FL model in a distributed and heterogeneous environment. In this survey, we first introduce the fundamentals of FUL. Then, we review the FUL frameworks that are proposed to address the three main implementation challenges, i.e., communication cost, resource allocation as well as security and privacy. Furthermore, we discuss applications of FUL in the modern distributed computer networks. We also highlight the open challenges and future research opportunities. By consolidating existing knowledge and mapping open problems, this survey aims to serve as a foundational reference for researchers and practitioners seeking to advance FL to build trustworthy, regulation-compliant and user-centric federated systems.

Figures (10)

• Figure 1: Structure of Federated Unlearning Survey.
• Figure 2: General framework of Federated Learning.
• Figure 3: Different types of Federated Unlearning.
• Figure 4: Achieving statistical indistinguishability between the retrained and unlearned models.
• Figure 5: Communication challenges of federated unlearning in mobile edge networks. The figure depicts a hierarchical deployment with mobile and IoT clients connected to edge servers and a cloud FL coordinator, and highlights how unlearning requests introduce extra communication paths on top of standard training rounds. Additional client–edge–cloud exchanges for deletion requests, rollback updates, and model resynchronization stress bandwidth constrained uplinks, amplify the impact of device churn and intermittent connectivity, and make naive retrain-based FUL impractical in real mobile edge environments.

Theorems & Definitions (1)

• Definition 1