Privacy of federated QR decomposition using additive secure multiparty computation

TL;DR

This paper explores privacy-preserving federated QR decomposition methods, proposing a secure Gram-Schmidt-based algorithm suitable for cross-silo federated learning, enabling privacy-aware linear regression without raw data leakage.

Contribution

It introduces a novel privacy-aware QR decomposition scheme based on Gram-Schmidt for federated learning, addressing privacy concerns in cross-silo settings.

Findings

Proposed a secure Gram-Schmidt QR decomposition algorithm.

Demonstrated the algorithm's application to federated linear regression.

Analyzed the privacy-preserving properties of the method.

Abstract

Federated learning (FL) is a privacy-aware data mining strategy keeping the private data on the owners' machine and thereby confidential. The clients compute local models and send them to an aggregator which computes a global model. In hybrid FL, the local parameters are additionally masked using secure aggregation, such that only the global aggregated statistics become available in clear text, not the client specific updates. Federated QR decomposition has not been studied extensively in the context of cross-silo federated learning. In this article, we investigate the suitability of three QR decomposition algorithms for cross-silo FL and suggest a privacy-aware QR decomposition scheme based on the Gram-Schmidt algorithm which does not blatantly leak raw data. We apply the algorithm to compute linear regression in a federated manner.