Secure Collaborative Training and Inference for XGBoost
This addresses privacy and security concerns for organizations in collaborative machine learning, though it is incremental as it builds on existing enclave technology with specific algorithmic enhancements.
The paper tackled the problem of enabling secure collaborative training and inference for XGBoost models by proposing Secure XGBoost, a system that protects data privacy and computation integrity using hardware enclaves and novel data-oblivious algorithms, resulting in a solution that prevents access side-channel attacks.
In recent years, gradient boosted decision tree learning has proven to be an effective method of training robust models. Moreover, collaborative learning among multiple parties has the potential to greatly benefit all parties involved, but organizations have also encountered obstacles in sharing sensitive data due to business, regulatory, and liability concerns. We propose Secure XGBoost, a privacy-preserving system that enables multiparty training and inference of XGBoost models. Secure XGBoost protects the privacy of each party's data as well as the integrity of the computation with the help of hardware enclaves. Crucially, Secure XGBoost augments the security of the enclaves using novel data-oblivious algorithms that prevent access side-channel attacks on enclaves induced via access pattern leakage.