Stochastic Client Selection for Federated Learning With Volatile Clients

Huang, Tiansheng; Lin, Weiwei; Li, Keqin; Zomaya, Albert Y.

doi:10.1109/jiot.2022.3172113

Cited by 64 publications

(29 citation statements)

References 13 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…"> K‐Center 29 : K‐Center clustering is performed on the set of clients before each training round to select the clients in each class that is closest to the center of clustering. E3CS 11 : Considering that clients cannot meet the deadline for training and the aggregation fails, E3CS generates the client's state (success or failure) using Bernoulli distribution, which is selected based on the expectation of the client's state.…”

Section: Methodsmentioning

confidence: 99%

“…However, during the footprint comparison process, FedProf requires uploading data feature information, which runs the risk of leaking the client's local data. Our team discovered in previous research 10,11 that low priority clients are denied participation in training during the client selection process for FL. Because of the inequity in the selection, the global model may be unable to effectively use data from some clients, and thus cannot guarantee data diversity on the global model aggregation, and lower global model performance.…”

Section: Related Workmentioning

confidence: 99%

“…Recently, some researchers have attempted to address the problem of unfair client selection and large bias in data distribution by assigning fairness factors to clients to ensure that all clients have a certain probability of being selected 10,11 . Experimental results show that the fairness‐based client selection algorithm can reduce the risk of overfitting while ensuring training efficiency when appropriate fairness values are chosen.…”

Section: Introductionmentioning

confidence: 99%

“…A client selection algorithm based on the contribution is proposed, inspired by the theory of Adversarial Bandit in Multiarmed Bandit (MAB) 12 . The client selection algorithm in this paper updates the client selection weights by estimating the contribution of the client, solving the problem of blindly imposing fairness constraints on the client while ignoring the actual contribution of the client in fairness client selection, and the approach in this paper (from the perspective of the contribution is more comprehensive than the approach in the literature 11 (only considers the case of client failure). It also does not affect the privacy‐preserving mechanism of FL.…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Contribution‐based Federated Learning client selection

Lin

Liu

et al. 2022

Int J of Intelligent Sys

Self Cite

View full text Add to dashboard Cite

Federated Learning (FL), as a privacy-preserving machine learning paradigm, has been thrusted into the limelight. As a result of the physical bandwidth constraint, only a small number of clients are selected for each round of FL training. However, existing client selection solutions (e.g., the vanilla random selection) typically ignore the heterogeneous data value of the clients. In this paper, we propose the contribution-based selection algorithm (Contribution-Based Exponentialweight algorithm for Exploration and Exploitation, CBE3), which dynamically updates the selection weights according to the impact of clients' data. As a novel component of CBE3, a scaling factor, which helps maintain a good balance between global model accuracy and convergence speed, is proposed to improve the algorithm's adaptability. Theoretically, we proved the regret bound of the proposed CBE3 algorithm, which demonstrates performance gaps between the CBE3 and the optimal choice. Empirically, extensive experiments conducted on Non-Independent Identically Distributed data demonstrate the superior performance of CBE3with up to 10% accuracy improvement compared with K-Center and Greedy and up to 100% faster convergence compared with the Random algorithm.

show abstract

Section: Methodsmentioning

confidence: 99%

Section: Related Workmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Contribution‐based Federated Learning client selection

Lin

Liu

et al. 2022

Int J of Intelligent Sys

Self Cite

View full text Add to dashboard Cite

show abstract

“…Therefore, we propose two approaches to identify malicious updates from different attacks (i.e., sybil and non-sybil), based on which rewards can be easily computed. Note that MAB methods have been applied in FL in other works, but they focus on reducing communication overheads [Cho et al, 2020] or constructing high-quality models [Huang et al, 2020], it is still unclear how to apply MAB algorithms to detect Byzantine attacks.…”

Section: Introductionmentioning

confidence: 99%

Shielding Federated Learning: Robust Aggregation with Adaptive Client Selection

Wan¹,

Hu²,

Lu³

et al. 2022

Preprint

View full text Add to dashboard Cite

Federated learning (FL) enables multiple clients to collaboratively train an accurate global model while protecting clients' data privacy. However, FL is susceptible to Byzantine attacks from malicious participants. Although the problem has gained significant attention, existing defenses have several flaws: the server irrationally chooses malicious clients for aggregation even after they have been detected in previous rounds; the defenses perform ineffectively against sybil attacks or in the heterogeneous data setting.To overcome these issues, we propose MAB-RFL, a new method for robust aggregation in FL. By modelling the client selection as an extended multi-armed bandit (MAB) problem, we propose an adaptive client selection strategy to choose honest clients that are more likely to contribute highquality updates. We then propose two approaches to identify malicious updates from sybil and nonsybil attacks, based on which rewards for each client selection decision can be accurately evaluated to discourage malicious behaviors. MAB-RFL achieves a satisfying balance between exploration and exploitation on the potential benign clients. Extensive experimental results show that MAB-RFL outperforms existing defenses in three attack scenarios under different percentages of attackers.

show abstract

A review on client selection models in federated learning

Panigrahi,

Bharti,

Sharma

2023

WIREs Data Min & Knowl

View full text Add to dashboard Cite

Federated learning (FL) is a decentralized machine learning (ML) technique that enables multiple clients to collaboratively train a common ML model without them having to share their raw data with each other. A typical FL process involves (1) FL client(s) selection, (2) global model distribution, (3) local training, and (4) aggregation. As such FL clients are heterogeneous edge devices (i.e., mobile phones) that differ in terms of computational resources, training data quality, and distribution. Therefore, FL client(s) selection has a significant influence on the execution of the remaining steps of an FL process. There have been a variety of FL client(s) selection models proposed in the literature, however, their critical review and/or comparative analysis is much less discussed. This paper brings the scattered FL client(s) selection models onto a single platform by first categorizing them into five categories, followed by providing a detailed analysis of the benefits/shortcomings and the applicability of these models for different FL scenarios. Such understanding can help researchers in academia and industry to develop improved FL client(s) selection models to address the requirement challenges and shortcomings of the current models. Finally, future research directions in the area of FL client(s) selection are also discussed.This article is categorized under: Technologies > Machine Learning Technologies > Artificial Intelligence

show abstract

Stochastic Client Selection for Federated Learning With Volatile Clients

Cited by 64 publications

References 13 publications

Contribution‐based Federated Learning client selection

Contribution‐based Federated Learning client selection

Shielding Federated Learning: Robust Aggregation with Adaptive Client Selection

A review on client selection models in federated learning

Contact Info

Product

Resources

About