Constant-Round Outsourcing Private Decision Tree Evaluation

Decision trees are widely deployed in critical applications such as medical diagnosis and financial analysis; however, their evaluation introduces significant privacy risks to both model parameters and client data. While Private Decision Tree Evaluation (PDTE) protocols offer privacy guarantees, existing solutions suffer from high latency due to multi-round communication. We propose an efficient outsourced PDTE protocol based on a three-server architecture that combines replicated secret sharing with function secret sharing. Our approach achieves a constant number of online rounds—regardless of tree depth—through optimized cryptographic operations, enabling efficient evaluation of sparse trees. Experiments show that our method achieves 1.86–2.85× faster online execution and maintains competitive communication efficiency compared to state-of-the-art approaches.