Abstract:In the rapidly expanding field of IoT, data production has reached an unprecedented scale, providing valuable insights that accelerate decision-making processes. However, ensuring the privacy and security of this massive amount of data poses significant challenges. In this paper, we propose using clustered federated learning (CFL) as a solution to ensure both the security and privacy of big data by uploading model weights while keeping the data stored locally. Nevertheless, there are practical challenges in applying CFL to big data: (1) the participating FL clients are unlikely to have identical data distributions; (2) insufficient attention is given to the similarity between different clusters; and (3) CFL tends to ignore the class imbalance problem (i.e., long-tailed), which hinders its application in big data and affects the quality of target tasks. To address these issues and enable widespread CFL deployment in big data applications, this paper proposes a prototype-assisted clustered federated learning framework (MDSPFL). It relaxes the assumption of unique data distribution for each client, allowing the client’s local dataset to follow multiple source distributions considering classification class imbalance, thereby aligning with clients in a big data environment. Specifically, MDSPFL employs the proximal update mechanism to handle workload surges caused by mixed distribution and unavailability of similarity between cluster models. Additionally, MDSPFL introduces a class-balanced local training mechanism to resolve the long-tailed problem, which utilizes contrastive learning and class prototypes to enforce a uniform distribution of all classes in the feature space. We conduct extensive experiments on different datasets (EMNIST, Cifar10, Cifar100), and the experimental results demonstrate the effectiveness of our proposed MDSPFL in big data scenarios with imbalance and mixed-distribution clients.

Taking Advantage of the Mistakes: Rethinking Clustered Federated Learning for IoT Anomaly Detection

Towards Fast and Accurate Federated Learning with Non-Iid Data for Cloud-Based IoT Applications

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

Hierarchical Federated Learning with Adaptive Clustering on Non-IID Data

CFL-IDS: an Effective Clustered Federated Learning Framework for Industrial Internet of Things Intrusion Detection

Geographical Node Clustering and Grouping to Guarantee Data IIDness in Federated Learning

LayerCFL: an Efficient Federated Learning with Layer-Wised Clustering

Clustered Data Sharing for Non-IID Federated Learning over Wireless Networks

ASCFL: Accurate and Speedy Semi-Supervised Clustering Federated Learning

Security and Privacy-Enhanced Federated Learning for Anomaly Detection in IoT Infrastructures

Adaptive Client Clustering for Efficient Federated Learning over Non-IID and Imbalanced Data

Federated Semi-Supervised and Semi-Asynchronous Learning for Anomaly Detection in IoT Networks

Stochastic Clustered Federated Learning

Deep Anomaly Detection for Time-Series Data in Industrial IoT: A Communication-Efficient On-Device Federated Learning Approach

Information-Exchangeable Hierarchical Clustering for Federated Learning with Non-IID Data

Energy-efficient Clustering to Address Data Heterogeneity in Federated Learning

A Cluster-Driven Adaptive Training Approach for Federated Learning

Behave Differently when Clustering: A Semi-asynchronous Federated Learning Approach for IoT

Flexible Clustered Federated Learning for Client-Level Data Distribution Shift

A Clustering Strategy for Enhanced FL-Based Intrusion Detection in IoT Networks

Fuzzy Clustered Federated Learning under Mixed Data Distributions.