Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Free-Rider and Conflict Aware Collaboration Formation for Cross-Silo Federated Learning

Mengmeng Chen, Xiaohu Wu, Xiaoli Tang, Tiantian He, Yew-Soon Ong, Qiqi Liu, Qicheng Lao, Han Yu

TL;DR

This work tackles cross-silo Federated Learning under dual forces of self-interest and inter-organization competition. It introduces a graph-theoretic framework with a benefit graph, a competition graph, and a data-usage graph to encode data complementarity, rivalry, and data sharing constraints, and formalizes two principles to avoid free-riding and conflicts of interest. The FedEgoists algorithm blends Bron–Kerbosch clique detection, Tarjan SCC decomposition, and a cycle/path/node-merging procedure to produce a coalition structure that is both stable and coalition-proof, with complexity dominated by the Bron–Kerbosch step. The approach demonstrates substantial performance gains over nine baselines on CIFAR-10/100 and a real-world eICU dataset, while preserving feasible collaboration topologies in cross-silo FL. Overall, the method offers a scalable, provably optimal way to form cross-silo FL coalitions that align with participants’ incentives and competitive dynamics, enabling more effective and trustworthy federated collaborations.

Abstract

Federated learning (FL) is a machine learning paradigm that allows multiple FL participants (FL-PTs) to collaborate on training models without sharing private data. Due to data heterogeneity, negative transfer may occur in the FL training process. This necessitates FL-PT selection based on their data complementarity. In cross-silo FL, organizations that engage in business activities are key sources of FL-PTs. The resulting FL ecosystem has two features: (i) self-interest, and (ii) competition among FL-PTs. This requires the desirable FL-PT selection strategy to simultaneously mitigate the problems of free riders and conflicts of interest among competitors. To this end, we propose an optimal FL collaboration formation strategy -- FedEgoists -- which ensures that: (1) a FL-PT can benefit from FL if and only if it benefits the FL ecosystem, and (2) a FL-PT will not contribute to its competitors or their supporters. It provides an efficient clustering solution to group FL-PTs into coalitions, ensuring that within each coalition, FL-PTs share the same interest. We theoretically prove that the FL-PT coalitions formed are optimal since no coalitions can collaborate together to improve the utility of any of their members. Extensive experiments on widely adopted benchmark datasets demonstrate the effectiveness of FedEgoists compared to nine state-of-the-art baseline methods, and its ability to establish efficient collaborative networks in cross-silos FL with FL-PTs that engage in business activities.

Free-Rider and Conflict Aware Collaboration Formation for Cross-Silo Federated Learning

TL;DR

This work tackles cross-silo Federated Learning under dual forces of self-interest and inter-organization competition. It introduces a graph-theoretic framework with a benefit graph, a competition graph, and a data-usage graph to encode data complementarity, rivalry, and data sharing constraints, and formalizes two principles to avoid free-riding and conflicts of interest. The FedEgoists algorithm blends Bron–Kerbosch clique detection, Tarjan SCC decomposition, and a cycle/path/node-merging procedure to produce a coalition structure that is both stable and coalition-proof, with complexity dominated by the Bron–Kerbosch step. The approach demonstrates substantial performance gains over nine baselines on CIFAR-10/100 and a real-world eICU dataset, while preserving feasible collaboration topologies in cross-silo FL. Overall, the method offers a scalable, provably optimal way to form cross-silo FL coalitions that align with participants’ incentives and competitive dynamics, enabling more effective and trustworthy federated collaborations.

Abstract

Federated learning (FL) is a machine learning paradigm that allows multiple FL participants (FL-PTs) to collaborate on training models without sharing private data. Due to data heterogeneity, negative transfer may occur in the FL training process. This necessitates FL-PT selection based on their data complementarity. In cross-silo FL, organizations that engage in business activities are key sources of FL-PTs. The resulting FL ecosystem has two features: (i) self-interest, and (ii) competition among FL-PTs. This requires the desirable FL-PT selection strategy to simultaneously mitigate the problems of free riders and conflicts of interest among competitors. To this end, we propose an optimal FL collaboration formation strategy -- FedEgoists -- which ensures that: (1) a FL-PT can benefit from FL if and only if it benefits the FL ecosystem, and (2) a FL-PT will not contribute to its competitors or their supporters. It provides an efficient clustering solution to group FL-PTs into coalitions, ensuring that within each coalition, FL-PTs share the same interest. We theoretically prove that the FL-PT coalitions formed are optimal since no coalitions can collaborate together to improve the utility of any of their members. Extensive experiments on widely adopted benchmark datasets demonstrate the effectiveness of FedEgoists compared to nine state-of-the-art baseline methods, and its ability to establish efficient collaborative networks in cross-silos FL with FL-PTs that engage in business activities.

Paper Structure

This paper contains 35 sections, 7 theorems, 12 equations, 12 figures, 7 tables, 5 algorithms.

Table of Contents

  1. Introduction
  2. Related Work
  3. Model and Assumptions
  4. Relationship Graphs
  5. Collaboration Principles
  6. Absence of free riders
  7. Avoiding conflict of interest
  8. Problem Description
  9. Solution
  10. Evaluation
  11. Experimental Setup
  12. Benchmark Experiments: CIFAR-10 & CIFAR-100
  13. Real-world Collaboration Example: eICU
  14. Conclusions
  15. Theoretical Analysis
  16. ...and 20 more sections

Key Result

Proposition 1

Upon completion of Algorithm algorithm1, Principles prin_free and prin_con are realized.

Figures (12)

  • Figure 1: An overview of the main motivation and results of this paper.
  • Figure 2: $\left\{SCC_{h}\right\}_{h=1}^{2}$ where $H=2$.
  • Figure 3: $\pi$
  • Figure 5: Label distribution of CIFAR-10. Colors indicate the labels for each FL-PT/FL-PT.
  • Figure 6: Label distribution of CIFAR-100. Colors indicate the labels for each FL-PT/FL-PT.
  • ...and 7 more figures

Theorems & Definitions (9)

  • Definition 1: Coalitions
  • Definition 2
  • Proposition 1
  • Proposition 2
  • Lemma 1
  • Lemma 2
  • Lemma 3
  • Lemma 4
  • Lemma 5