Accelerating Vertical Federated Learning

Privacy, security and data governance constraints rule out a brute force process in the integration of cross-silo data, which inherits development Internet Things. Federated learning is proposed to ensure that all parties can collaboratively complete training task while not local. Vertical federated specialization for distributed features. To preserve privacy, homomorphic encryption applied enable encrypted operations without decryption. Nevertheless, together with robust guarantee, brings extra communication computation overhead. In this paper, we analyze current bottlenecks vertical under comprehensively numerically. We propose straggler-resilient computation-efficient accelerating system reduces overhead heterogeneous scenarios by 65.26% at most caused 40.66% most. Our improve robustness efficiency framework loss security.