Optimizing Secure Decision Tree Inference Outsourcing

Outsourcing decision tree inference services to the cloud is highly beneficial, yet raises critical privacy concerns on proprietary of model provider and private input data client. In this paper, we design, implement, evaluate a new system that allows efficient outsourcing inference. Our significantly improves upon prior art in overall online end-to-end secure service latency at as well local-side performance provider. We first present scheme which securely shifts most processing cloud, resulting substantial reduction provider's complexities. further devise substantially optimizes for particularly communication round The synergy these techniques our achieve up $8 \times$ better side over realistic WAN environment, bring notation="LaTeX">$19 savings notation="LaTeX">$18 computation.