Boosting Communication Efficiency of Federated Learning's Secure Aggregation

TL;DR

This paper introduces CESA, a communication-efficient secure aggregation protocol for federated learning that reduces overhead while maintaining privacy, especially in stable networks with low delay variation.

Contribution

It presents a novel CESA protocol that lowers communication costs in federated learning's secure aggregation, building on Google's SecAgg method.

Findings

CESA significantly reduces communication overhead compared to SecAgg.

CESA maintains privacy by preventing server access to unmasked models.

The protocol is effective in stable, low-delay networks with limited client dropouts.

Abstract

Federated Learning (FL) is a decentralized machine learning approach where client devices train models locally and send them to a server that performs aggregation to generate a global model. FL is vulnerable to model inversion attacks, where the server can infer sensitive client data from trained models. Google's Secure Aggregation (SecAgg) protocol addresses this data privacy issue by masking each client's trained model using shared secrets and individual elements generated locally on the client's device. Although SecAgg effectively preserves privacy, it imposes considerable communication and computation overhead, especially as network size increases. Building upon SecAgg, this poster introduces a Communication-Efficient Secure Aggregation (CESA) protocol that substantially reduces this overhead by using only two shared secrets per client to mask the model. We propose our method for stable networks with low delay variation and limited client dropouts. CESA is independent of the data distribution and network size (for higher than 6 nodes), preventing the honest-but-curious server from accessing unmasked models. Our initial evaluation reveals that CESA significantly reduces the communication cost compared to SecAgg.