BAFFLE : Blockchain Based Aggregator Free Federated Learning

A key aspect of Federated Learning (FL) is the requirement of a centralized aggregator to store and update the global model. However, in many cases orchestrating a centralized aggregator might be infeasible due to numerous operational constraints. In this paper, we introduce BAFFLE, an aggregator free, blockchain driven, FL environment that is inherently decentralized. BAFFLE leverages Smart Contracts (SC) to store the global copy of the model, delineate the FL mechanism into distinct rounds and aggregate local models and update the global copy after each round. BAFFLE boosts computational performance by first decomposing the global parameter space into distinct chunks followed by a novel score and bid strategy leading to significant reduction in computational costs on the blockchain. In order to validate our claims we conduct extensive experiments using a pertinent case study on a private Ethereum network and demonstrate the computational efficiency and scalability of BAFFLE. Further, our results also show that BAFFLE delivers similar performance as its centralized as well as classical FL counterparts in addition to minimizing the computational overhead of blockchain based decentralization.