Federated Learning using Peer-to-peer Network for Decentralized
Orchestration of Model Weights
Abstract
In recent times, Machine learning and Artificial intelligence have
become one of the key emerging fields of computer science. Many
researchers and businesses are benefited by machine learning models that
are trained by data processing at scale. However, machine learning, and
particularly Deep Learning requires large amounts of data, that in
several instances are proprietary and confidential to many businesses.
In order to respect individual organization’s privacy in collaborative
machine learning, federated learning could play a crucial role. Such
implementations of privacy preserving federated learning find
applicability in various ecosystems like finance, health care, legal,
research and other fields that require preservation of privacy. However,
many such implementations are driven by a centralized architecture in
the network, where the aggregator node becomes the single point of
failure, and is also expected with lots of computing resources at its
disposal. In this paper, we propose an approach of implementing a
decentralized, peer-topeer federated learning framework, that leverages
RAFT based aggregator selection. The proposal hinges on that fact that
there is no one permanent aggregator, but instead a transient, time
based elected leader, which will aggregate the models from all the peers
in the network. The leader ( aggregator) publishes the aggregated model
on the network, for everyone to consume. Along with peer-to-peer network
and RAFT based aggregator selection, the framework uses dynamic
generation of cryptographic keys, to create a more secure mechanism for
delivery of models within the network. The key rotation also ensures
anonymity of the sender on the network too. Experiments conducted in the
paper, verifies the usage of peer-to-peer network for creating a
resilient federated learning network. Although the proposed solution
uses an artificial neural network in it’s reference implementation, the
generic design of the framework can accommodate any federated learning
model within the network.