Intent
Define a four-layer privacy architecture where Data, Execution, Settlement, and Disclosure are distinct concerns, each addressable by different technologies. This enables institutions to select best-fit components per layer while maintaining interoperability and upgrade paths.Best for: Different privacy requirements exist for different transaction phases, or organization needs flexibility to swap components without full redesign.
Architecture Overview
Ingredients
Standards
- ERC-7573: Atomic DvP settlement across networks
- ERC-3643: Compliant security tokens with identity hooks
- EAS: Ethereum Attestation Service for disclosure proofs
- EIP-4844: Blob transactions for data availability
Infrastructure by Layer
| Layer | Responsibility | Options |
|---|---|---|
| Data | Store/retrieve encrypted data | Off-chain encrypted storage, DA layers (EigenDA, Celestia), L2 blobs (EIP-4844), IPFS+encryption |
| Execution | Private computation | ZK (Aztec, Miden), FHE (Zama, Fhenix), TEE enclaves (Flashbots SUAVE) |
| Settlement | Finality, atomicity | Ethereum L1, Arbitrum, Optimism, ZKsync, Polygon |
| Disclosure | Regulatory access | View keys, ZK proofs, threshold KMS, EAS attestations |
Off-chain Services
- Key management systems per layer (data encryption keys, proving keys, signing keys)
- Cross-layer routing and orchestration (transaction coordinator)
- Audit log aggregation (unified view across layers)
Protocol
Define Layer Boundaries
Map each transaction phase to a layer:Example: Private bond issuance
| Phase | Layer | Technology |
|---|---|---|
| Store bond terms | Data | Encrypted off-chain storage |
| Verify investor eligibility | Execution | ZK proof on Aztec |
| Transfer ownership | Settlement | ERC-7573 on Ethereum L1 |
| Regulator access | Disclosure | View keys via threshold KMS |
Route Data to Storage
Encrypt sensitive transaction data and store via the Data layer:Returns content-addressed reference (
contentHash) for downstream layers.Execute Private Computation
The Execution layer retrieves encrypted data references, performs computation, and outputs verifiable state transition:
Settle on Target Chain
Submit execution proof to Settlement layer:Settlement may be on L1 for maximum security or L2 for lower cost.
Generate Disclosure Artifacts
Produce disclosure proofs, view keys, or attestations for authorized parties:
Guarantees
Layer Isolation
Layer Isolation
Compromise or failure in one layer does not expose data in others:
- Data layer breach: Encrypted blobs remain secure; attacker cannot decrypt without keys
- Execution layer failure: Settlement layer can revert; stored data unaffected
- Settlement layer censorship: Can migrate to alternative settlement chain without re-encrypting data
- Disclosure layer key leak: Only affects scoped disclosures; does not compromise execution or data encryption
Composability
Composability
Layers can be independently upgraded or swapped:Example upgrade paths:
- Migrate from TEE execution → ZK execution without changing settlement
- Move from L1 settlement → L2 settlement without changing data storage
- Add new disclosure mechanism (e.g., FHE-based queries) alongside existing view keys
- Switch from IPFS → Arweave for data storage without re-generating proofs
Auditability
Auditability
Cross-layer audit trail enables reconstruction of transaction flow:
Failure Containment
Failure Containment
Settlement layer defines atomicity boundary. If execution fails, settlement reverts without affecting stored data:
Trade-offs
Latency Accumulation
Each layer adds processing time:| Layer | Typical Latency | Cumulative |
|---|---|---|
| Data | 100-500ms (storage write) | 100-500ms |
| Execution | 5-30s (proof generation) | 5.1-30.5s |
| Settlement | 12-180s (block confirmation) | 17-210s |
| Disclosure | 1-5s (key generation) | 18-215s |
Total latency: ~18 seconds to ~3.5 minutes. Compare to monolithic solutions: ~30 seconds to 2 minutes.
Interface Rigidity
Standardized interfaces may constrain layer-specific optimizations:- Data layer: Generic encryption interface may not support FHE-specific optimizations
- Execution layer: Abstraction over ZK/FHE/TEE may limit access to advanced features
- Settlement layer: ERC-7573 may not cover all atomic settlement patterns
Tooling Fragmentation
Different vendors per layer means multiple SDKs, monitoring tools, and support relationships:Metadata Leakage
Even with encrypted data, cross-layer routing can reveal timing, access patterns, and transaction graph:- Timing correlation: Data write → execution proof → settlement transaction reveals trade flow
- Size leakage: Encrypted blob sizes may hint at transaction complexity
- Access patterns: Frequent retrieval from specific content hashes reveals hot data
- Add random delays between layer transitions
- Pad encrypted blobs to fixed sizes
- Use mixing services or batch processing to obscure individual transactions
Example: Institutional Bond Settlement
Data Layer: Store Bond Terms
Bond terms and investor details encrypted with AES-256, stored in off-chain encrypted storage:Content hash
0x7f3a... published to Celestia DA layer for availability guarantees.Execution Layer: Compute Allocation
Aztec private contracts compute allocation and verify investor eligibility via ZK proofs:Execution produces settlement instruction with proof of valid state transition.
Settlement Layer: Atomic DvP on L1
ERC-7573 atomic DvP on Ethereum L1:Transaction hash
0xdef456... recorded on Ethereum L1 at block 18,456,789.Disclosure Layer: Regulator Access
Issuer provides view keys to BaFin via EAS attestation:Regulator can:
- Decrypt bond terms and investor details using view key
- Verify investor eligibility proofs from execution layer
- Confirm settlement on L1 block explorer
- Other bond issuances (scoped view key)
- Full transaction graph (only specific trade)
- Investor details in other trades
Upgrade Scenario
This modular configuration allows the issuer to:- Upgrade Execution Layer
- Add New Regulator
- Move to L2 Settlement
Migrate from Aztec to Miden for execution without changing settlement:
- Replace Aztec client with Miden VM client
- Rewrite circuits in Miden assembly
- Data layer, settlement layer, disclosure layer unchanged
- Zero downtime migration via parallel deployment
Implementation Checklist
Define Layer Requirements
- Identify data sensitivity levels
- Determine execution privacy needs (ZK/FHE/TEE)
- Select settlement finality requirements (L1/L2)
- Define disclosure scope and authorized parties
Select Layer Technologies
- Choose data storage backend (IPFS/Arweave/Celestia)
- Choose execution environment (Aztec/Zama/Flashbots)
- Choose settlement chain (Ethereum/Arbitrum/Polygon)
- Choose disclosure mechanism (view keys/ZK proofs/EAS)
Implement Layer Interfaces
- Standardize API contracts between layers
- Implement cross-layer orchestration logic
- Add error handling and retry mechanisms
- Build unified audit logging
Test Integration
- Unit test each layer independently
- Integration test cross-layer flows
- Stress test coordination under load
- Security audit layer boundaries
See Also
Private L2s
Execution layer options: Aztec, Miden, Scroll
Commit and Prove
Cross-layer coordination pattern
Atomic DvP via ERC-7573
Settlement layer standard
Regulatory Disclosure
Disclosure layer mechanisms
L2 Encrypted Off-chain Audit
Data + disclosure combination pattern
TEE-based Privacy
Execution layer option using trusted hardware
Vendor Implementations
Aztec
ZK execution layer
Zama
FHE execution layer
Miden
ZK VM execution layer
Railgun
Privacy infrastructure
EigenDA
Data availability layer
Celestia
Modular DA network

