Crate atlas_c2pa_lib 

Atlas C2PA Library

atlas-c2pa-lib is a Rust library for creating, signing, and verifying machine learning assets (models and datasets) with C2PA (Content Provenance and Authenticity) specifications.

The library provides tools to generate cryptographic claims about ML asset provenance, track ML asset lineage, and create C2PA-compliant manifests.

Key Components

• Assertions: Claims about ML models and datasets
• Ingredients: Tracking datasets and components used to create models
• Manifests: Complete C2PA manifests for ML assets
• Asset Types: Support for various ML frameworks (TensorFlow, PyTorch, ONNX, etc.)

Example: Creating a Model Manifest

use atlas_c2pa_lib::ml::types::{ModelInfo, MLFramework, ModelFormat};
use atlas_c2pa_lib::ml::manifest::MLManifestBuilder;
use time::OffsetDateTime;

// Define model information
let model_info = ModelInfo {
    name: "bert-base".to_string(),
    version: "1.0.0".to_string(),
    framework: MLFramework::PyTorch,
    format: ModelFormat::TorchScript,
};

// Create a manifest builder
let builder = MLManifestBuilder::new(
    model_info.clone(),
    "0123456789abcdef0123456789abcdef".to_string(), // Model hash
);

// Build the manifest
let manifest = builder.build().unwrap();

Modules

assertion

Assertion Module

asset_type

Asset Type Module

cbor

CBOR Module

claim

Claim Module

cose

COSE Module

cross_reference

Cross Reference Module

datetime_wrapper

DateTime Wrapper Module

ingredient

Ingredient Module

manifest

Manifest Module

ml

ML Module