Search papers, labs, and topics across Lattice.
This paper introduces a privacy-preserving vertical federated learning (VFL) framework with three protocols tailored for different deployment scenarios, addressing both input and output privacy. The approach distributes the aggregator role across multiple servers that use secure multiparty computation (MPC) for model and feature aggregation, and differential privacy (DP) for the final model. By optimizing the computation and communication via MPC, particularly through global-local model updates, the proposed framework significantly reduces the overhead compared to naive MPC-based VFL.
Vertical federated learning can be made practical and private with a new framework that drastically cuts MPC overhead by distributing aggregation and using global-local model updates.
We propose a novel end-to-end privacy-preserving framework, instantiated by three efficient protocols for different deployment scenarios, covering both input and output privacy, for the vertically split scenario in federated learning (FL), where features are split across clients and labels are not shared by all parties. We do so by distributing the role of the aggregator in FL into multiple servers and having them run secure multiparty computation (MPC) protocols to perform model and feature aggregation and apply differential privacy (DP) to the final released model. While a naive solution would have the clients delegating the entirety of training to run in MPC between the servers, our optimized solution, which supports purely global and also global-local models updates with privacy-preserving, drastically reduces the amount of computation and communication performed using multiparty computation. The experimental results also show the effectiveness of our protocols.