Crate weirwood 

weirwood — privacy-preserving XGBoost inference via Fully Homomorphic Encryption.

Load a trained XGBoost model and evaluate it either in plaintext (for testing) or encrypted under FHE so the server learns nothing about the input.

Quickstart

use weirwood::{model::WeirwoodTree, eval::{Evaluator, PlaintextEvaluator}};

let weirwood_tree = WeirwoodTree::from_json_file("model.json")?;
let features = vec![1.0_f32, 0.5, 3.2, 0.1];
let score = PlaintextEvaluator.predict(&weirwood_tree, &features);

Re-exports

pub use error::Error;

Modules

error

eval

Evaluators for running inference over a loaded WeirwoodTree.

fhe

model

XGBoost model loading and internal representation.