ZKNeuralSwiftCore

ZKNeuralSwiftCore is a Swift package that provides a high-level interface for generating zero-knowledge proofs (ZKPs) for neural network computations. It leverages a Rust core via FFI and supports invoking TensorFlow Lite models and generating ZKPs for their outputs.

Features

• Zero-Knowledge Proof Generation: Easily generate ZKPs for neural network inference results.
• Tensor Invoker: Run TFLite models and obtain outputs in a format suitable for proof generation.
• Rust Core Integration: Swift bindings to a performant Rust backend.
• Callbacks: Customizable witness and proof generation via Swift closures.

Usage

Add ZKNeuralSwiftCore as a dependency in your Package.swift using Swift Package Manager:

.package(url: "https://github.com/rarimo/ZKNeuralSwiftCore.git", from: "0.1.0")

Example: Generating a ZK Proof

import ZKNeuralSwiftCore

let config = ZKNeuralCoreConfiguration(
    generateWitnessCallback: { /* your witness callback */ },
    generateProofCallback: { /* your proof callback */ },
    provingType: ZKNeuralProvingTypeOptions.UltraGroth
)
let core = ZKNeuralCore(configuration: config)
let proof = try core.generateZkProof(inputJson, circuit, zkey)

Example: Invoking a Tensor Model

let invoker = TensorInvoker(modelData)
let output = try invoker.fireImage(imageData)

Example: Generating Generic Inputs for Circuits

import ZKNeuralSwiftCore

let invoker = TensorInvoker(modelData)

let inputs = try invoker.drainGenericInputs("3123", "3123", "312", false, imageData, options: ImagePreprocessing::FaceRecognition) 

License

MIT License. See LICENCE for details.