concrete-core

Crates.ioconcrete-core
lib.rsconcrete-core
version1.0.2
sourcesrc
created_at2021-03-25 16:38:24.211041
updated_at2022-11-30 08:24:30.834785
descriptionConcrete is a fully homomorphic encryption (FHE) library that implements Zama's variant of TFHE.
homepagehttps://www.zama.ai/concrete-framework
repositoryhttps://github.com/zama-ai/concrete-core
max_upload_size
id373453
size6,612,412
Zama Bot (zama-bot)

documentation

https://docs.zama.ai/concrete-core

README

Concrete Core

This crate contains low-level implementations of homomorphic operators used in the concrete library (GitHub Repo).

⚠ Warning ⚠

This crate assumes that the user is comfortable with the theory behind FHE. If you prefer to use a simpler API, that will perform sanity checks on your behalf, the higher-level concrete crate should have your back.

Example

Here is a small example of how one could use concrete-core to perform a simple operation homomorphically:

// This examples shows how to multiply a secret value by a public one homomorphically.

// First we import the proper symbols
use concrete_core::prelude::Variance;
use concrete_core::prelude::LweDimension;
use concrete_core::prelude::*;
use std::error::Error;

fn main() -> Result<(), Box<dyn Error>> {
    // DISCLAIMER: the parameters used here are only for test purpose, and cannot be considered secure.
    let lwe_dimension = LweDimension(750);
    let noise = Variance(2_f64.powf(-104.));

    // Here a hard-set encoding is applied on the input (shift by 59 bits) which corresponds here
    // to a precision of 4 bits with an additional bit of padding (won't be used but required for
    // PBS)
    let raw_input = 3_u64 << 59;

    // We will multiply by 4
    let raw_input_cleatext = 4_u64;

    // Unix seeder must be given a secret input.
    // Here we just give it 0, which is totally unsafe.
    const UNSAFE_SECRET: u128 = 0;
    let mut engine = DefaultEngine::new(Box::new(UnixSeeder::new(UNSAFE_SECRET)))?;

    // We create a cleartext from the raw cleartext
    let cleartext: Cleartext64 = engine.create_cleartext_from(&raw_input_cleatext)?;
    let key: LweSecretKey64 = engine.generate_new_lwe_secret_key(lwe_dimension)?;

    // We crate the input plaintext from the raw input
    let input_plaintext = engine.create_plaintext_from(&raw_input)?;
    let input_ciphertext = engine.encrypt_lwe_ciphertext(&key, &input_plaintext, noise)?;

    // Create a container for the output, whose content will be discarded during the operation
    let mut output_ciphertext =
        engine.trivially_encrypt_lwe_ciphertext(lwe_dimension.to_lwe_size(), &input_plaintext)?;

    // Perform the multiplication, overwriting (discarding) the output ciphertext content
    engine.discard_mul_lwe_ciphertext_cleartext(
        &mut output_ciphertext,
        &input_ciphertext,
        &cleartext
    )?;

    // Get the decrypted result as a plaintext and then a raw value
    let decrypted_plaintext = engine.decrypt_lwe_ciphertext(&key, &output_ciphertext)?;
    let raw_decrypted_plaintext = engine.retrieve_plaintext(&decrypted_plaintext)?;

    // Round the output for our 4 bits of precision
    let output = raw_decrypted_plaintext >> 58;
    let carry = output % 2;
    let output = ((output >> 1) + carry) % (1 << 5);

    // Check the high bits have the result we expect
    assert_eq!(output, 12);

    Ok(())
}

Links

License

This software is distributed under the BSD-3-Clause-Clear license. If you have any questions, please contact us at hello@zama.ai.

Commit count: 720

cargo fmt