self_encryption

Self encrypting files (convergent encryption plus obfuscation)

Crate	Documentation

Table of Contents

• Overview
• Documentation
• Features
• Usage
  • Rust Usage
    • Basic Operations
    • Storage Backends
    • Streaming Operations
    • Advanced Usage
  • Python Usage
• Implementation Details
• License
• Contributing
• Release Process

Overview

A version of convergent encryption with an additional obfuscation step. This pattern allows secured data that can also be de-duplicated. This library presents an API that takes a set of bytes and returns a secret key derived from those bytes, and a set of encrypted chunks.

Important Security Note: While this library provides very secure encryption of the data, the returned secret key requires the same secure handling as would be necessary for any secret key.

Documentation

• Self Encrypting Data Whitepaper
• Process Overview Video

Features

• Content-based chunking
• Convergent encryption
• Self-validating chunks
• Hierarchical data maps for handling large files
• Streaming encryption/decryption
• Python bindings
• Flexible storage backend support
• Custom storage backends via functors

Usage

Rust Usage

Installation

Add this to your Cargo.toml:

[dependencies]
self_encryption = "0.30"
bytes = "1.0"

Basic Operations

use self_encryption::{encrypt, decrypt_full_set};
use bytes::Bytes;

// Basic encryption/decryption
fn basic_example() -> Result<()> {
    let data = Bytes::from("Hello, World!".repeat(1000));  // Must be at least 3072 bytes
    
    // Encrypt data
    let (data_map, encrypted_chunks) = encrypt(data.clone())?;
    
    // Decrypt data
    let decrypted = decrypt(&data_map, &encrypted_chunks)?;
    assert_eq!(data, decrypted);
    
    Ok(())
}

Storage Backends

use self_encryption::{shrink_data_map, get_root_data_map, decrypt_from_storage};
use std::collections::HashMap;
use std::sync::{Arc, Mutex};

// Memory Storage Example
fn memory_storage_example() -> Result<()> {
    let storage = Arc::new(Mutex::new(HashMap::new()));
    
    // Store function
    let store = |hash, data| {
        storage.lock().unwrap().insert(hash, data);
        Ok(())
    };
    
    // Retrieve function
    let retrieve = |hash| {
        storage.lock().unwrap()
            .get(&hash)
            .cloned()
            .ok_or_else(|| Error::Generic("Chunk not found".into()))
    };
    
    // Use with data map operations
    let shrunk_map = shrink_data_map(data_map, store)?;
    let root_map = get_root_data_map(shrunk_map, retrieve)?;
    
    Ok(())
}

// Disk Storage Example
fn disk_storage_example() -> Result<()> {
    let chunk_dir = PathBuf::from("chunks");
    
    // Store function
    let store = |hash, data| {
        let path = chunk_dir.join(hex::encode(hash));
        std::fs::write(path, data)?;
        Ok(())
    };
    
    // Retrieve function
    let retrieve = |hash| {
        let path = chunk_dir.join(hex::encode(hash));
        Ok(Bytes::from(std::fs::read(path)?))
    };
    
    // Use with data map operations
    let shrunk_map = shrink_data_map(data_map, store)?;
    let root_map = get_root_data_map(shrunk_map, retrieve)?;
    
    Ok(())
}

Python Usage

Installation

pip install self-encryption

Basic Operations

from self_encryption import encrypt, decrypt

# Basic in-memory encryption/decryption
def basic_example():
    # Create test data (must be at least 3072 bytes)
    data = b"Hello, World!" * 1000
    
    # Encrypt data - returns data map and encrypted chunks
    data_map, chunks = encrypt(data)
    print(f"Data encrypted into {len(chunks)} chunks")
    print(f"Data map has child level: {data_map.child()}")
    
    # Decrypt data
    decrypted = decrypt(data_map, chunks)
    assert data == decrypted

Implementation Details

Core Process

• Files are split into chunks of up to 1MB
• Each chunk is processed in three steps:
  1. Compression (using Brotli)
  2. Encryption (using AES-256-CBC)
  3. XOR obfuscation

Key Generation and Security

• Each chunk's encryption uses keys derived from the content hashes of three chunks:

  For chunk N:
  - Uses hashes from chunks [N, N+1, N+2]
  - Combined hash = hash(N) || hash(N+1) || hash(N+2)
  - Split into:
    - Pad (first X bytes)
    - Key (next 16 bytes for AES-256)
    - IV  (final 16 bytes)
  

• This creates a chain of dependencies where each chunk's encryption depends on its neighbors

• Provides both convergent encryption and additional security through the interdependencies

Encryption Flow

1. Content Chunking:

   • File is split into chunks of optimal size
   • Each chunk's raw content is hashed (SHA3-256)
   • These hashes become part of the DataMap

2. Per-Chunk Processing:

   // For each chunk:
   1. Compress data using Brotli
   2. Generate key materials:
      - Combine three consecutive chunk hashes
      - Extract pad, key, and IV
   3. Encrypt compressed data using AES-256-CBC
   4. XOR encrypted data with pad for obfuscation
   

3. DataMap Creation:

   • Stores both pre-encryption (src) and post-encryption (dst) hashes
   • Maintains chunk ordering and size information
   • Required for both encryption and decryption processes

Decryption Flow

1. Chunk Retrieval:

   • Use DataMap to identify required chunks
   • Retrieve chunks using dst_hash as identifier

2. Per-Chunk Processing:

   // For each chunk:
   1. Regenerate key materials using src_hashes from DataMap
   2. Remove XOR obfuscation using pad
   3. Decrypt using AES-256-CBC with key and IV
   4. Decompress using Brotli
   

3. Chunk Reassembly:

   • Chunks are processed in order specified by DataMap
   • Reassembled into original file

Storage Features

• Flexible backend support through trait-based design

• Supports both memory and disk-based storage

• Streaming operations for memory efficiency

• Hierarchical data maps for large files:

  // DataMap shrinking for large files
  1. Serialize large DataMap
  2. Encrypt serialized map using same process
  3. Create new DataMap with fewer chunks
  4. Repeat until manageable size reached
  

Security Properties

• Content-based convergent encryption
• Additional security through chunk interdependencies
• Self-validating chunks through hash verification
• No single point of failure in chunk storage
• Tamper-evident through hash chains

Performance Optimizations

• Parallel chunk processing where possible
• Streaming support for large files
• Efficient memory usage through chunking
• Optimized compression settings
• Configurable chunk sizes

This implementation provides a balance of:

• Security (through multiple encryption layers)
• Deduplication (through convergent encryption)
• Performance (through parallelization and streaming)
• Flexibility (through modular storage backends)

License

Licensed under the General Public License (GPL), version 3 (LICENSE http://www.gnu.org/licenses/gpl-3.0.en.html).

Linking Exception

self_encryption is licensed under GPLv3 with linking exception. This means you can link to and use the library from any program, proprietary or open source; paid or gratis. However, if you modify self_encryption, you must distribute the source to your modified version under the terms of the GPLv3.

See the LICENSE file for more details.

Contributing

Want to contribute? Great :tada:

There are many ways to give back to the project, whether it be writing new code, fixing bugs, or just reporting errors. All forms of contributions are encouraged!

For instructions on how to contribute, see our Guide to contributing.

Release Process

To prepare a new release:

1. Create a PR with version bump and changelog:

   • Update version in Cargo.toml based on Semantic Versioning
   • Add new version entry to CHANGELOG.md with release date and changes
   • Example: PR #416

2. Run the release workflow manually:

   • After PR is merged, go to GitHub Actions
   • Click "Run workflow" to trigger the automated release process