Robust Server Defense Against Unreliable Clients in One-Shot Fair Collaborative Machine Learning

2026-05-08

AI Summary

This work addresses the challenge of fairness degradation in single-round collaborative learning caused by biased proxy data submitted by unreliable clients, which disproportionately harms underrepresented groups. The paper proposes the first server-side defense framework that dynamically learns client weights through a bilevel optimization formulation, leveraging a small trusted root dataset to mitigate bias and enforce fairness constraints within a single communication round. Notably, the method operates without iterative training and remains effective even when the majority of clients are untrustworthy. Empirical evaluations on standard benchmarks demonstrate that the approach significantly improves model fairness while incurring minimal accuracy loss, outperforming existing methods.

Abstract

Collaborative machine learning (CML) enables multiple clients to train a global model jointly in a data-distributed setting. To address data privacy and communication efficiency, one-shot CML has been increasingly adopted, where clients communicate with the server only once by sharing synthetic or processed proxy data. This single-round communication, however, eliminates the possibility of iterative correction at the server, making the learning process particularly vulnerable to client unreliability. In this setting, unreliable clients, whether malicious or non-malicious, may provide biased proxy data that favors certain groups, thereby degrading the fairness of the global model and harming minority or unprivileged groups. In this work, we propose a server-side defense framework based on a bilevel optimization formulation. The proposed approach learns client-level weights to mitigate the influence of biased client proxy data while enforcing fairness constraints by using a very small trusted root dataset available at the server. Experimental results on benchmark datasets show that our method improves fairness with little accuracy loss under biased proxy data contributions from unreliable clients. Moreover, the proposed approach remains effective even when unreliable clients make up a majority of the system, consistently outperforming other existing methods.

Problem

• collaborative machine learning
• one-shot learning
• client unreliability
• model fairness
• proxy data bias

Innovation

• one-shot collaborative learning
• fairness
• bilevel optimization
• client unreliability
• trusted root dataset