Trustless Federated Learning at Edge-Scale: A Compositional Architecture for Decentralized, Verifiable, and Incentive-Aligned Coordination

A comprehensive empirical study has been conducted on the privacy vulnerabilities of Spiking Neural Networks (SNNs) in the context of Federated Learning (FL), particularly focusing on gradient leakage attacks. This research highlights the potential for sensitive training data to be reconstructed from shared gradients, a concern that has been extensively studied in conventional Artificial Neural Networks (ANNs) but remains largely unexplored in SNNs.

A recent study has established the first benchmark for applying differential privacy in federated learning for automatic speech recognition, addressing challenges associated with training large transformer models. The research highlights the issues of gradient heterogeneity and proposes techniques such as per-layer clipping and layer-wise gradient normalization to improve convergence rates.