Abstract:—A traditional federated learning (FL) system allows clients to collaboratively train a global model under the coordination of a central server, which sparks great interests in exploiting the private data distributed on clients’ local devices. However, once the central server suffers from a single point of failure, it will lead to system crash. In addition, the FL system usually involves a large number of clients, which requires expensive communication costs. These challenges inspire a communication-efﬁcient design of decentralized FL systems. In this paper, we propose an efﬁcient and privacy-preserving global model training protocol in the context of FL in large-scale peer-to-peer networks, CFL. The proposed CFL protocol aggregates local model update parameters hierarchically by a cluster-based aggregation mode, as well as a leverged authenticated encryption scheme to ensure the security communication, whose key is distributed by a modiﬁed secure communication key establishment protocol. Theoretical analyses show that the proposed CFL protocol guarantees the privacy of local model update parameters, as well as integrity and authenticity under the widespread internal semi-honest and external malicious threat models. In particular, the proposed key revocation based on public voting can effectively defense against external adversaries hijacking honest participants to ensure the conﬁdentiality of the communication keys. In addition, the modiﬁed secure communication key establishment protocol indeed achieves high network connectivity probability to ensure transmission security of the system. Moreover, experimental results on the Trec06p and Trec07 datasets show that the ﬁnal global model trained through the proposed CFL protocol can achieve an accuracy of 99.32% after 14 training rounds, which is comparable to the accuracy and convergence speed of the PPT protocol. The proposed CFL protocol is equally excellent even facing dropout clients. Compared with the PPT protocol, the CFL protocol improves communication efﬁciency by 43.25% and computational efﬁciency by 0.87%.

ClassTer: Mobile Shift-Robust Personalized Federated Learning Via Class-Wise Clustering

Hierarchical Federated Learning with Adaptive Clustering on Non-IID Data

Flexible Clustered Federated Learning for Client-Level Data Distribution Shift

EchoPFL: Asynchronous Personalized Federated Learning on Mobile Devices with On-Demand Staleness Control

DistFL: Distribution-aware Federated Learning for Mobile Scenarios

FedReMa: Improving Personalized Federated Learning via Leveraging the Most Relevant Clients

Communication-Efficient and Model-Heterogeneous Personalized Federated Learning via Clustered Knowledge Transfer

Cooperative Multi-Model Training for Personalized Federated Learning over Heterogeneous Devices

CFL: Cluster Federated Learning in Large-scale Peer-to-Peer Networks

Hierarchical Personalized Federated Learning Over Massive Mobile Edge Computing Networks

A Comprehensive Study on Personalized Federated Learning with Non-IID Data.

Towards Hierarchical Clustered Federated Learning with Model Stability on Mobile Devices

ContextFL: Context-aware Federated Learning by Estimating the Training and Reporting Phases of Mobile Clients

Group Personalized Federated Learning

A Prototype-Assisted Clustered Federated Learning for Big Data Security and Privacy Preservation

FedClust: Tackling Data Heterogeneity in Federated Learning through Weight-Driven Client Clustering

Communication-Efficient Cluster Federated Learning in Large-scale Peer-to-Peer Networks.

Accelerating Federated Learning Via Parameter Selection and Pre-Synchronization in Mobile Edge-Cloud Networks

Personalized Federated Learning with Model Interpolation among Client Clusters and Its Application in Smart Home

CPPer-FL: Clustered Parallel Training for Efficient Personalized Federated Learning

A Personalized Federated Learning Framework Using Side Information for Heterogeneous Data Classification.