Reproduce “One-Shot Secure Aggregation: A Hybrid Cryptographic Protocol for Private Federated Learning in IoT” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Imraul Emmaka, Tran Viet Xuan Phuong
A hybrid cryptographic protocol, Hyb-Agg, enables one-shot secure aggregation for federated learning in IoT settings, achieving constant per-client communication regardless of the number of participants. The University of Arkansas at Little Rock research demonstrates sub-second execution times on a Raspberry Pi 4, with client processing taking 431.8 ms and server aggregation 191.1 ms for 50 clients and 8192 model parameters.
Key insights
- The protocol's hybrid design leverages MK-CKKS for homomorphic addition across encrypted updates and ECDH masking for securing partial decryption shares, enabling one-shot aggregation.
- The specific construction of ECDH-based additive masks ensures that their sum across all clients is zero, allowing the server to correctly recover the aggregate.
- Parameter selection for CKKS, focusing on additive depth and scaling factors, achieves numerical exactness for aggregated floating-point values while meeting 128-bit security requirements.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.