Since Teaclave is a FaaS-like platform, users can define and register customized functions (e.g., written in Python). To make data computation more easier and faster (in native speed), the platform also provide some commonly used functions written in Rust. We call them built-in functions. These functions can be selectively compiled in the built-in executor with a “builtin” prefix in the function names. Users can selectively invoke these build-in functions by their names.
Currently, we have these built-in functions:
builtin-echo: Return the original input message.
builtin-gbdt-train: Use input data to train a GBDT model.
builtin-gbdt-predict: GBDT prediction with input model and input test data.
bulitin-logistic-regression-train: Use input data to train a LR model.
builtin-logistic-regression-predict: LR prediction with input model and input test data.
builtin-private-join-and-compute: Find intersection of muti-parties' input data and compute sum of the common items.
builtin-ordered-set-intersect: Allow two parties to compute the intersection of their ordered sets without revealing anything except for the elements in the intersection. Users should calculate hash values of each item and upload them as a sorted list.
builtin-rsa-sign: Signing data with RSA key.
builtin-face-detection: An implementation of Funnel-Structured cascade, which is designed for real-time multi-view face detection.
builtin-principal-components-analysis: Example to calculate PCA.
builtin-password-check: Given a password, check whether it is in the exposed password list.
The function arguments are in JSON format and can be serialized to a Rust struct very easily. You can learn more about supported arguments in the implementation of a specific built-in function.