Abstract:We propose a novel framework, , for multiparty computations, with full threshold and semi-honest security model, supporting a combination of real number arithmetic (arithmetic shares), Boolean arithmetic (Boolean shares) and garbled circuits (Yao shares). In contrast to prior work (Mohassel and Zhang, in 2017 IEEE symposium on security and privacy (SP), 2017; Mohassel and Rindal, in Proceedings of the 2018 ACM SIGSAC conference on computer and communications security, 2018), mitigates overflows, which is of paramount importance for machine learning applications, without compromising efficiency or security. Compared to other overflow-free recent techniques such as MP-SPDZ (Escudero et al., in 40th annual international cryptology conference, CRYPTO. Lecture notes in computer science, 2020) that convert arithmetic to Boolean shares, uses an efficient modular lifting/truncation method that allows for scalable high numerical precision computations with optimal numerical windows and hence, highly efficient online phases. We adapt basic MPC operations such as real-valued polynomial evaluation, division, logarithms, exponentials, Fourier series evaluations and oblivious comparisons to by employing our modular lift in combination with existing efficient conversions between arithmetic, Boolean and Yao shares. We also describe a highly scalable computations of logistic regression models with real-world training data sizes and high numerical precision through PCA and blockwise variants (for memory and runtime optimizations) based on second-order optimization techniques. On a dataset of 50 M samples and 50 features distributed among two players, the online phase completes in 14.5 h with at least 10 decimal digits of precision compared to plaintext training. The setup phase of is supported in both the trusted dealer and the interactive models allowing for tradeoffs between efficiency and stronger security. The highly efficient online phase makes the framework particularly suitable for MPC applications where the output of the setup phase is part of the input of the protocol (such as MPC-in-the-head or Prio ).

XORBoost: Tree Boosting in the Multiparty Computation Setting

A Hybrid-Domain Framework for Secure Gradient Tree Boosting.

Scalable Multi-Party Privacy-Preserving Gradient Tree Boosting over Vertically Partitioned Dataset with Outsourced Computations

Secure Collaborative Training and Inference for XGBoost

SoK: Modular and Efficient Private Decision Tree Evaluation.

BOOST: Block Minifloat-Based On-Device CNN Training Accelerator with Transfer Learning

E-Booster: An FPGA-based Accelerator for Secure Tree Boosting Using Additively Homomorphic Encryption

SGBoost: An Efficient and Privacy-Preserving Vertical Federated Tree Boosting Framework

SilentWood: Private Inference Over Gradient-Boosting Decision Forests

Optimization of Oblivious Decision Tree Ensembles Evaluation for CPU

OpBoost

Towards End-to-End Secure and Efficient Federated Learning for XGBoost

OpBoost: A Vertical Federated Tree Boosting Framework Based on Order-Preserving Desensitization

Efficient decision tree training with new data structure for secure multi-party computation

Federated Boosted Decision Trees with Differential Privacy

SecureBoost+: Large Scale and High-Performance Vertical Federated Gradient Boosting Decision Tree

Securely Training Decision Trees Efficiently

Secure training of decision trees with continuous attributes

GTree: GPU-Friendly Privacy-preserving Decision Tree Training and Inference

Manticore: A Framework for Efficient Multiparty Computation Supporting Real Number and Boolean Arithmetic

SecureBoost: A Lossless Federated Learning Framework