aitbc/docs/reference/confidential-transactions.md

# Confidential Transactions Architecture

## Overview

Design for opt-in confidential transaction support in AITBC, enabling participants to encrypt sensitive transaction data while maintaining selective disclosure and audit capabilities.

## Architecture

### Encryption Model

**Hybrid Encryption with Envelope Pattern**:
1. **Data Encryption**: AES-256-GCM for transaction data
2. **Key Exchange**: X25519 ECDH for per-recipient key distribution
3. **Envelope Pattern**: Random DEK per transaction, encrypted for each authorized party

### Key Components

```
┌─────────────────┐    ┌──────────────────┐    ┌─────────────────┐
│   Transaction   │───▶│  Encryption      │───▶│  Storage        │
│   Service       │    │  Service         │    │  Layer          │
└─────────────────┘    └──────────────────┘    └─────────────────┘
         │                       │                       │
         ▼                       ▼                       ▼
┌─────────────────┐    ┌──────────────────┐    ┌─────────────────┐
│   Key Manager   │    │  Access Control  │    │  Audit Log      │
└─────────────────┘    └──────────────────┘    └─────────────────┘
```

## Data Flow

### 1. Transaction Creation (Opt-in)

```python
# Client requests confidential transaction
transaction = {
    "job_id": "job-123",
    "amount": "1000",
    "confidential": True,
    "participants": ["client-456", "miner-789", "auditor-001"]
}

# Coordinator encrypts sensitive fields
encrypted = encryption_service.encrypt(
    data={"amount": "1000", "pricing": "details"},
    participants=transaction["participants"]
)

# Store with encrypted payload
stored_transaction = {
    "job_id": "job-123",
    "public_data": {"job_id": "job-123"},
    "encrypted_data": encrypted.ciphertext,
    "encrypted_keys": encrypted.encrypted_keys,
    "confidential": True
}
```

### 2. Data Access (Authorized Party)

```python
# Miner requests access to transaction data
access_request = {
    "transaction_id": "tx-456",
    "requester": "miner-789",
    "purpose": "settlement"
}

# Verify access rights
if access_control.verify(access_request):
    # Decrypt using recipient's private key
    decrypted = encryption_service.decrypt(
        ciphertext=stored_transaction.encrypted_data,
        encrypted_key=stored_transaction.encrypted_keys["miner-789"],
        private_key=miner_private_key
    )
```

### 3. Audit Access (Regulatory)

```python
# Auditor with court order requests access
audit_request = {
    "transaction_id": "tx-456",
    "requester": "auditor-001",
    "authorization": "court-order-123"
}

# Special audit key escrow
audit_key = key_manager.get_audit_key(audit_request.authorization)
decrypted = encryption_service.audit_decrypt(
    ciphertext=stored_transaction.encrypted_data,
    audit_key=audit_key
)
```

## Implementation Details

### Encryption Service

```python
class ConfidentialTransactionService:
    """Service for handling confidential transactions"""
    
    def __init__(self, key_manager: KeyManager):
        self.key_manager = key_manager
        self.cipher = AES256GCM()
    
    def encrypt(self, data: Dict, participants: List[str]) -> EncryptedData:
        """Encrypt data for multiple participants"""
        # Generate random DEK
        dek = os.urandom(32)
        
        # Encrypt data with DEK
        ciphertext = self.cipher.encrypt(dek, json.dumps(data))
        
        # Encrypt DEK for each participant
        encrypted_keys = {}
        for participant in participants:
            public_key = self.key_manager.get_public_key(participant)
            encrypted_keys[participant] = self._encrypt_dek(dek, public_key)
        
        # Add audit escrow
        audit_public_key = self.key_manager.get_audit_key()
        encrypted_keys["audit"] = self._encrypt_dek(dek, audit_public_key)
        
        return EncryptedData(
            ciphertext=ciphertext,
            encrypted_keys=encrypted_keys,
            algorithm="AES-256-GCM+X25519"
        )
    
    def decrypt(self, ciphertext: bytes, encrypted_key: bytes, 
                private_key: bytes) -> Dict:
        """Decrypt data for specific participant"""
        # Decrypt DEK
        dek = self._decrypt_dek(encrypted_key, private_key)
        
        # Decrypt data
        plaintext = self.cipher.decrypt(dek, ciphertext)
        return json.loads(plaintext)
```

### Key Management

```python
class KeyManager:
    """Manages encryption keys for participants"""
    
    def __init__(self, storage: KeyStorage):
        self.storage = storage
        self.key_pairs = {}
    
    def generate_key_pair(self, participant_id: str) -> KeyPair:
        """Generate X25519 key pair for participant"""
        private_key = X25519.generate_private_key()
        public_key = private_key.public_key()
        
        key_pair = KeyPair(
            participant_id=participant_id,
            private_key=private_key,
            public_key=public_key
        )
        
        self.storage.store(key_pair)
        return key_pair
    
    def rotate_keys(self, participant_id: str):
        """Rotate encryption keys"""
        # Generate new key pair
        new_key_pair = self.generate_key_pair(participant_id)
        
        # Re-encrypt active transactions
        self._reencrypt_transactions(participant_id, new_key_pair)
```

### Access Control

```python
class AccessController:
    """Controls access to confidential transaction data"""
    
    def __init__(self, policy_store: PolicyStore):
        self.policy_store = policy_store
    
    def verify_access(self, request: AccessRequest) -> bool:
        """Verify if requester has access rights"""
        # Check participant status
        if not self._is_authorized_participant(request.requester):
            return False
        
        # Check purpose-based access
        if not self._check_purpose(request.purpose, request.requester):
            return False
        
        # Check time-based restrictions
        if not self._check_time_restrictions(request):
            return False
        
        return True
    
    def _is_authorized_participant(self, participant_id: str) -> bool:
        """Check if participant is authorized for confidential transactions"""
        # Verify KYC/KYB status
        # Check compliance flags
        # Validate regulatory approval
        return True
```

## Data Models

### Confidential Transaction

```python
class ConfidentialTransaction(BaseModel):
    """Transaction with optional confidential fields"""
    
    # Public fields (always visible)
    transaction_id: str
    job_id: str
    timestamp: datetime
    status: str
    
    # Confidential fields (encrypted when opt-in)
    amount: Optional[str] = None
    pricing: Optional[Dict] = None
    settlement_details: Optional[Dict] = None
    
    # Encryption metadata
    confidential: bool = False
    encrypted_data: Optional[bytes] = None
    encrypted_keys: Optional[Dict[str, bytes]] = None
    algorithm: Optional[str] = None
    
    # Access control
    participants: List[str] = []
    access_policies: Dict[str, Any] = {}
```

### Access Log

```python
class ConfidentialAccessLog(BaseModel):
    """Audit log for confidential data access"""
    
    transaction_id: str
    requester: str
    purpose: str
    timestamp: datetime
    authorized_by: str
    data_accessed: List[str]
    ip_address: str
    user_agent: str
```

## Security Considerations

### 1. Key Security
- Private keys stored in HSM or secure enclave
- Key rotation every 90 days
- Zero-knowledge proof of key possession

### 2. Data Protection
- AES-256-GCM provides confidentiality + integrity
- Random IV per encryption
- Forward secrecy with per-transaction DEKs

### 3. Access Control
- Multi-factor authentication for decryption
- Role-based access control
- Time-bound access permissions

### 4. Audit Compliance
- Immutable audit logs
- Regulatory access with court orders
- Privacy-preserving audit proofs

## Performance Optimization

### 1. Lazy Encryption
- Only encrypt fields marked as confidential
- Cache encrypted data for frequent access
- Batch encryption for bulk operations

### 2. Key Management
- Pre-compute shared secrets for regular participants
- Use key derivation for multiple access levels
- Implement key caching with secure eviction

### 3. Storage Optimization
- Compress encrypted data
- Deduplicate common encrypted patterns
- Use column-level encryption for databases

## Migration Strategy

### Phase 1: Opt-in Support
- Add confidential flags to existing models
- Deploy encryption service
- Update transaction endpoints

### Phase 2: Participant Onboarding
- Generate key pairs for all participants
- Implement key distribution
- Train users on privacy features

### Phase 3: Full Rollout
- Enable confidential transactions by default for sensitive data
- Implement advanced access controls
- Add privacy analytics and reporting

## Testing Strategy

### 1. Unit Tests
- Encryption/decryption correctness
- Key management operations
- Access control logic

### 2. Integration Tests
- End-to-end confidential transaction flow
- Cross-system key exchange
- Audit trail verification

### 3. Security Tests
- Penetration testing
- Cryptographic validation
- Side-channel resistance

## Compliance

### 1. GDPR
- Right to encryption
- Data minimization
- Privacy by design

### 2. Financial Regulations
- SEC Rule 17a-4
- MiFID II transaction reporting
- AML/KYC requirements

### 3. Industry Standards
- ISO 27001
- NIST Cybersecurity Framework
- PCI DSS for payment data

## Next Steps

1. Implement core encryption service
2. Create key management infrastructure
3. Update transaction models and APIs
4. Deploy access control system
5. Implement audit logging
6. Conduct security testing
7. Gradual rollout with monitoring