Boosting Communication Efficiency of Federated Learning's Secure Aggregation

TL;DR

The paper addresses privacy-preserving aggregation in Federated Learning by tackling the high communication overhead of Google's SecAgg. It introduces CESA, a protocol that uses two shared masks per client derived from a public-key exchange to mask updates, eliminating encryption and per-pair masking. The key insight is that pairwise masks cancel out in aggregation, making server communication largely independent of network size for networks with at least 6 nodes. Empirical evaluation across 100 rounds shows that CESA significantly reduces communication costs compared to SecAgg, offering scalable secure aggregation suitable for stable 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.

Figures (3)

• Figure 1: Comparison between Vanilla FL and FL with SecAgg.
• Figure 2: SecAgg (left) vs. CESA (right): pair selection and shared secret creation.
• Figure 3: Transmitted messages between clients and the server.