DPBalance: Efficient and Fair Privacy Budget Scheduling for Federated Learning as a Service
Yu Liu, Zibo Wang, Yifei Zhu, Chen Chen
TL;DR
This work addresses privacy budget scheduling in Federated Learning as a Service (FLaaS) by modeling privacy loss as a non-replenishable resource and jointly optimizing efficiency and fairness. It introduces DPBalance, a sequential allocation framework that uses a Lagrange-multiplier-based decomposition plus greedy heuristics to maximize a platform utility that blends dominant efficiency and fairness, controlled by parameters $\beta$ and $\lambda$. The authors prove four fundamental economic properties (Pareto Efficiency, Sharing Incentive, Envy-Freeness, and Strategy Proofness) and demonstrate a theoretical fairness–efficiency tradeoff, with empirical results showing substantial gains over state-of-the-art baselines in both efficiency and fairness. This approach enables practical, principled DP budgeting in FLaaS, improving throughput and equitable data-analyst participation across diverse pipelines.
Abstract
Federated learning (FL) has emerged as a prevalent distributed machine learning scheme that enables collaborative model training without aggregating raw data. Cloud service providers further embrace Federated Learning as a Service (FLaaS), allowing data analysts to execute their FL training pipelines over differentially-protected data. Due to the intrinsic properties of differential privacy, the enforced privacy level on data blocks can be viewed as a privacy budget that requires careful scheduling to cater to diverse training pipelines. Existing privacy budget scheduling studies prioritize either efficiency or fairness individually. In this paper, we propose DPBalance, a novel privacy budget scheduling mechanism that jointly optimizes both efficiency and fairness. We first develop a comprehensive utility function incorporating data analyst-level dominant shares and FL-specific performance metrics. A sequential allocation mechanism is then designed using the Lagrange multiplier method and effective greedy heuristics. We theoretically prove that DPBalance satisfies Pareto Efficiency, Sharing Incentive, Envy-Freeness, and Weak Strategy Proofness. We also theoretically prove the existence of a fairness-efficiency tradeoff in privacy budgeting. Extensive experiments demonstrate that DPBalance outperforms state-of-the-art solutions, achieving an average efficiency improvement of $1.44\times \sim 3.49 \times$, and an average fairness improvement of $1.37\times \sim 24.32 \times$.