Data-Free Contribution Estimation in Federated Learning using Gradient von Neumann Entropy

2026-04-24

AI Summary

This work addresses the lack of reliable, data-free methods for evaluating client contributions in federated learning without requiring validation data or client metadata—a gap that leaves existing approaches vulnerable to privacy leakage or manipulation. The authors propose, for the first time, using the von Neumann spectral entropy of clients’ final-layer gradient update matrices as a data-free contribution signal that captures the diversity of information provided by each client. Building on this insight, they design SpectralFed and SpectralFuse, which integrate normalized weighting, class alignment, and rank-adaptive Kalman filtering. Evaluated on non-IID benchmarks including CIFAR-10/100, FEMNIST, and FedISIC, the proposed entropy-based scores exhibit strong correlation with clients’ standalone accuracy and significantly outperform existing data-free baselines.

Technology Category

• Machine Learning: Distributed Machine Learning & Federated Learning
• Computer Vision: Adversarial Attacks & Robustness
• Constraint Satisfaction and Optimization: Distributed CSP/Optimization

Application Category

• User Modeling, Personalization and Recommendation: Federated recommendation systems and personalization
• Security and Privacy: Large-scale security measurements
• Responsible Web: Data and user privacy-enhancing technologies for the Web

Abstract

Client contribution estimation in Federated Learning is necessary for identifying clients' importance and for providing fair rewards. Current methods often rely on server-side validation data or self-reported client information, which can compromise privacy or be susceptible to manipulation. We introduce a data-free signal based on the matrix von Neumann (spectral) entropy of the final-layer updates, which measures the diversity of the information contributed. We instantiate two practical schemes: (i) SpectralFed, which uses normalized entropy as aggregation weights, and (ii) SpectralFuse, which fuses entropy with class-specific alignment via a rank-adaptive Kalman filter for per-round stability. Across CIFAR-10/100 and the naturally partitioned FEMNIST and FedISIC benchmarks, entropy-derived scores show a consistently high correlation with standalone client accuracy under diverse non-IID regimes - without validation data or client metadata. We compare our results with data-free contribution estimation baselines and show that spectral entropy serves as a useful indicator of client contribution.

Problem

• Federated Learning
• Client Contribution Estimation
• Data-Free
• Privacy
• Non-IID

Innovation

• von Neumann entropy
• data-free contribution estimation
• Federated Learning
• spectral entropy
• client importance