I was wondering if there exists a machine learning library designed to be run in an enclave. If not is it possible to use popular frameworks like Keras or Tensorflow by using Graphene-sgx ? Or dlib since it is a native C++ library ? Or should I implement my own algorithms ?
Thanks in advance and have a good day,
Update: Today we released gbdt-rs at https://github.com/mesalock-linux/gbdt-rs . It works well in SGX can directly inference by using a model exported by xgboost. Its paper is accepted at IEEE S&P 2019
Poster: gbdt-rs: Fast and Trustworthy Gradient Boosting Decision Tree
Tianyi Li (Peking University); Tongxin Li, Yu Ding, Yulong Zhang, and Tao Wei (Baidu X-Lab); Xinhui Han (Peking University)