Two-Tier Data Packing in RLWE-based Homomorphic Encryption for Secure Federated Learning

What is it about?

This paper proposes a novel data encoding method designed to improve the efficiency of Homomorphic Encryption (HE) in Federated Learning (FL). The primary issue it addresses is the computational and communication overhead that arises with HE, especially in the context of RLWE-based HE.

Featured Image

Why is it important?

Existing data encoding methods for HE in FL, such as simple concatenation of integers or polynomial coefficients, do not fully utilize the potential of HE based on Ring Learning with Errors (RLWE), leading to inefficiencies. This two-tier encoding strategy enhances batch processing for HE, making it more flexible and efficient for FL tasks. It provides improvements in computational and communication efficiency compared to existing batch HE methods.