Model CBOR Serialization for Federated Learning

TL;DR

This work addresses the challenge of performing federated learning on resource-constrained edge devices where JSON/gRPC messaging is impractical. It introduces TinyFL, a CBOR-based, CoAP-enabled message framework that reuses existing firmware components to enable efficient model dissemination and aggregation. Through extensive evaluation, including LeNet-5 and synthetic model sizes, the approach achieves up to $75\%$ reduction in serialized message size compared to JSON, with many updates fitting within single transmission frames. The results demonstrate substantial practical gains for FL on constrained networks, suggesting significant impact for IoT-scale distributed learning and easier integration with existing IoT infrastructures. The work also outlines future directions such as personalization and partial-transfer learning to further enhance performance under tight resource constraints.

Abstract

The typical federated learning workflow requires communication between a central server and a large set of clients synchronizing model parameters between each other. The current frameworks use communication protocols not suitable for resource-constrained devices and are either hard to deploy or require high-throughput links not available on these devices. In this paper, we present a generic message framework using CBOR for communication with existing federated learning frameworks optimised for use with resource-constrained devices and low power and lossy network links. We evaluate the resulting message sizes against JSON serialized messages where compare both with model parameters resulting in optimal and worst case serialization length, and with a real-world LeNet-5 model. Our benchmarks show that with our approach, messages are up to 75 % smaller in size when compared to the JSON alternative.

Figures (2)

• Figure 1: CoAP communication between FL Server and Clients.
• Figure 2: CoAP communication timing diagram between FL Server and a single client.