FLOSS: Federated Learning with Opt-Out and Straggler Support

2025-07-30

AI Summary

In federated learning, users’ exercise of data deletion rights and device heterogeneity jointly induce straggler nodes and data scarcity, degrading model performance and yielding biased parameter estimates. To address this, we propose FLOSS—a unified system that jointly models both sources of data deficiency. FLOSS integrates dynamic participation modeling, robust aggregation, and adaptive scheduling, while co-designing differential privacy guarantees and fault tolerance. Unlike conventional approaches that overlook the diversity of data scarcity and suffer significant performance degradation, FLOSS ensures model convergence and accuracy even under high client dropout rates (>50%) and severe stragglers (>10× mean latency), incurring only a 1.2% accuracy drop on CIFAR-10. By holistically reconciling privacy compliance, system heterogeneity, and statistical robustness, FLOSS substantially improves training stability and fairness in real-world privacy-sensitive deployments.

Technology Category

• Machine Learning: Distributed Machine Learning & Federated Learning
• Search and Optimization: Learning to Search
• Natural Language Processing: Safety and Robustness

Application Category

• User Modeling, Personalization and Recommendation: Federated recommendation systems and personalization
• Systems and Infrastructure for Web, Mobile and WoT: Federated Web and WoT systems, including distributed, federated and edge-based data processing
• Security and Privacy: Data transparency and provenance

Abstract

Previous work on data privacy in federated learning systems focuses on privacy-preserving operations for data from users who have agreed to share their data for training. However, modern data privacy agreements also empower users to use the system while opting out of sharing their data as desired. When combined with stragglers that arise from heterogeneous device capabilities, the result is missing data from a variety of sources that introduces bias and degrades model performance. In this paper, we present FLOSS, a system that mitigates the impacts of such missing data on federated learning in the presence of stragglers and user opt-out, and empirically demonstrate its performance in simulations.

Problem

• Handles missing data in federated learning due to user opt-out
• Addresses performance degradation from stragglers in heterogeneous devices
• Reduces bias caused by incomplete data sources in training

Innovation

• Federated learning with opt-out support
• Straggler mitigation in federated learning
• Bias reduction via missing data handling