Office of the Chief Digital Currency Architect
USD$ System Architecture
A technical overview of the digital dollar’s settlement, verification, and reserve-attestation infrastructure.
Public Technical Reference · Version 1.1 · Revised May 2031
Prepared for authorized intermediaries, reserve custodians, independent auditors, and security researchers.
1. Purpose and scope.
This document describes the production architecture of the United States digital dollar (USD$) and the infrastructure operated by the Federal Digital Dollar Corporation (the Corporation) under the Digital Dollar and Monetary Innovation Act of 2029. It is intended for parties who must integrate with, audit, or independently verify the system, and assumes familiarity with distributed systems and applied cryptography. It does not cover internal operational procedures, incident response, or information governed by separate confidentiality obligations.
2. Design principle.
The architecture is built on a single governing principle: the security of the system should depend on the smallest possible set of claims a user must take on faith. Rather than asking a holder of USD$ to trust the Corporation’s solvency and honesty in the abstract, the system is built so that its central claim — that every outstanding digital dollar is backed by an equivalent reserve asset — is independently and continuously verifiable by cryptographic proof, published for anyone to check.
This does not eliminate trust from the system. It narrows where trust is required and makes the remaining requirement explicit. Section 9 states that requirement precisely.
3. Architectural layers.
The system is organized into four layers, each with a distinct responsibility. The layers are logically separate even where they run on shared infrastructure.
4. Transaction lifecycle.
The layers in Section 3 are easiest to understand in motion. This section traces a single USD$ transfer from initiation to final settlement.
- Initiation. A customer instructs an authorized intermediary (Section 9) to transfer USD$ to a counterparty. The intermediary authenticates the customer and constructs a signed transfer request.
- Submission. The intermediary submits the request to the settlement ledger’s transaction gateway in the region currently authoritative for its shard (Section 8).
- Validation. The gateway checks the request’s signature, the sender’s available balance in the current ledger state, and applicable compliance holds, then forwards it into the validator set’s proposal queue.
- Consensus. A rotating proposer includes the transaction in a candidate block. Each validator independently verifies the block against its own copy of the ledger state and signs it if valid. The block is final once it carries signatures from at least two-thirds of the validator set, weighted by voting power (Section 5).
- Settlement. On finality, ledger balances update atomically. Settlement is final and irrevocable; there is no netting window, and no reversal path other than a separately authorized correction process, itself subject to the same threshold-signature requirement.
- Confirmation and mirroring. The ledger issues a settlement receipt to both intermediaries. Where the transfer involves the public circulation layer (Section 6), a corresponding mint or burn instruction, carrying the same validator signatures, is submitted to the relevant public-network bridge.
Under normal operation this sequence completes in low single-digit seconds from submission to final settlement receipt, bounded primarily by the consensus round in step four rather than network transit.
5. The permissioned settlement ledger.
The settlement ledger is not a permissionless, proof-of-work network, and it has no miners. Participation in consensus is restricted to a fixed, certified roster of identified, regulated institutional validators — a distinction the Corporation maintains carefully, since the security model of an anonymous open network does not transfer to a sovereign monetary system.
Consensus follows a Byzantine fault-tolerant protocol. A rotating proposer assembles each candidate block; the block finalizes once at least two-thirds of the validator set, weighted by voting power, has independently verified and signed it. This tolerates up to one-third of the validator set acting arbitrarily, including maliciously, without compromising the ledger’s correctness or liveness. As of launch the validator set numbers in the low dozens, drawn from Treasury, the Federal Reserve, the Corporation, and a set of large, qualifying financial institutions certified under Title IV of the Act.
Validators are geographically and institutionally diverse. Cryptographic keys are held in hardware security modules under threshold authorization, so that no single validator, administrator, or facility can unilaterally alter monetary state. Keys are rotated on a fixed schedule and immediately upon suspected compromise; rotation itself requires threshold sign-off from the existing validator set, so that no single validator — including one about to be replaced — can act unilaterally on its own membership.
6. The USD$ token and circulation layer.
One USD$ is, at all times, redeemable on demand for one United States dollar. Its value derives from that redemption relationship and the reserves backing it, not from scarcity or market speculation. The token is mirrored from the settlement ledger onto public networks so it can circulate wherever dollar-denominated digital commerce occurs, including transactions initiated by automated and institutional counterparties operating without direct human action at each step.
Movement between the settlement ledger and a public network runs through a bridge contract deployed on that network. To mint USD$ publicly, an intermediary first locks an equivalent balance on the settlement ledger; once that lock is final, the validator set signs a mint authorization, and the bridge contract accepts it only if the signatures satisfy the same two-thirds threshold described in Section 5, verified on-chain against the validator set’s published keys. Burning reverses the sequence: a holder submits a burn transaction on the public network, the bridge contract locks and marks it, and the validator set signs the corresponding unlock on the settlement ledger. The bridge contract holds no independent minting authority of its own — it verifies validator signatures and nothing more — so a compromise of a public network’s execution environment cannot create USD$ that the settlement ledger does not equally recognize. Canonical balances and settlement finality remain on the permissioned ledger at all times; the public networks that carry circulation are a reach layer, not a second seat of authority.
7. Reserve-verification layer: proof of reserves.
A reserve claim is only useful if it answers two questions together: how much the Corporation owes, and how much it holds. The verification layer binds these into a single publicly checkable proof.
Every outstanding USD$ liability is represented in a cryptographically committed liability set, constructed as a Merkle-sum tree. Each leaf commits to a blinded reference to one customer liability and its value, using a keyed commitment so that two leaves belonging to the same customer cannot be correlated by an outside observer; each parent commits to both the hashes and the sums of its children. The resulting root commits simultaneously to which liabilities exist, their aggregate value, and the structure of the set, without publicly disclosing any individual balance. A holder or their intermediary can request an inclusion proof — a short path of sibling hashes from their leaf to the published root — and verify it locally, without contacting the Corporation again.
Reserve assets are held by segregated, independent custodians. Each custodian produces a signed, timestamped attestation of its eligible balances at scheduled windows, twice each day, and each attestation must carry the sign-off of all three of that custodian’s designated signers, working from separate stations under separate operational control, before it is accepted into the verification layer. The liability commitment is bound to the most recent attested reserve state, and the resulting proof is republished every few minutes and anchored on a public network, where any party can independently verify it without relying on the Corporation’s own assurance.
Stated formally: let Lt be the digital dollars outstanding at time t, and Atatt the reserves attested at that time. The protocol continuously and publicly establishes Lt ≤ Atatt.
8. Deployment topology.
Ledger state is partitioned across a small number of geographic regions for latency and resilience, each region serving one partition of the outstanding supply. Each partition is replicated asynchronously to a paired standby region, which holds a copy of that partition’s ledger a few seconds behind the authoritative site, within a bounded recovery window. Failover to the standby is automated and tested against the loss of any single region. At any time, exactly one site is authoritative for a given partition and permitted to accept writes, which prevents two sites from independently accepting conflicting writes to the same state. The specific regions and their pairings are operational detail outside the scope of this public document.
The validator and authority plane described in Section 5 is logically and physically separate from the operational plane that serves a given region — the ledger store, transaction gateways, customer-facing interfaces, and the reserve-attestation oracle that carries custodians’ signed attestations into the verification layer. Because the ledger is hash-chained and independently signed by validators, a change in which region is actively serving a partition changes nothing about the ledger’s validated authenticity.
The oracle layer does not rely on a single attestation feed. Each custodian’s attestation arrives through an independent channel, and the oracle requires attestations from a quorum of custodians before publishing an updated reserve figure; an attestation that diverges materially from the others is flagged and excluded pending review rather than blended into the published figure. This bounds the effect a single compromised or mistaken attestation channel can have on the published figure. It does not remove the layer’s underlying reliance on custodians and signers acting honestly, which Section 10 states directly.
9. Authorized intermediaries.
Banks and other regulated financial institutions participate as authorized intermediaries: certified by the Corporation to distribute, hold, transfer, and redeem USD$ on behalf of customers, subject to digital-identity, know-your-customer, and anti-money-laundering standards set by the Corporation. Intermediaries connect through a certified gateway API, authenticated by mutual TLS and per-transaction message signatures; message formats follow ISO 20022 payment and reporting schemas wherever the underlying semantics permit, so intermediaries can extend existing correspondent-banking tooling rather than build a bespoke integration.
Certification is granted at one of three tiers under Title V of the Act: reporting-only access, for supervisory and audit purposes; standard access, for customer-facing transfer, redemption, and balance inquiry; and settlement access, reserved for institutions large and well-capitalized enough to also be considered for the validator role described in Section 5. Intermediary participation is separate from the validator role; certification at any tier is not itself validator status, though a subset of qualifying institutions may hold both. Certification requires ongoing compliance with the Corporation’s standards and can be suspended unilaterally on a documented compliance failure.
10. Trust assumptions and threat model.
The architecture is designed against a threat model that includes criminal organizations, insider actors, compromised institutions, hostile states, and coordinated cyber operations.
| Attack vector | Design mitigation | Residual |
|---|---|---|
| Anonymous consensus takeover | Permissioned, identified validator set; no open participation | Validator identity is itself an institutional trust point, addressed by regulatory oversight rather than cryptography |
| Single validator key compromise | Threshold signatures; hardware-backed keys; rotation on suspicion | Requires compromise of two-thirds of voting power to affect the ledger |
| Issuer concealing or inflating liabilities | Public, cryptographically committed liability tree; independently verifiable inclusion proofs | None, while the commitment scheme and hash function hold |
| False or stale reserve attestation | Three-signer attestations; multi-custodian quorum; outlier rejection; proof republished every few minutes | Requires coordinated compromise across independent custodians and signers within one attestation interval |
| Public-network bridge compromise | Bridge accepts only validator-threshold-signed instructions; holds no independent minting authority | Execution risk on a public network is contained to that network; it cannot mint supply the settlement ledger does not recognize |
One assumption cannot be removed by cryptography and is stated here directly rather than left implicit. The protocol establishes that Lt ≤ Atatt — that outstanding USD$ does not exceed the reserves as attested. It cannot, by itself, establish that the attestation accurately describes the physical reserves it reports. That determination rests on the custody-and-attestation layer: the independence and integrity of custodians, their signers, and the channel by which their attestations reach the verification layer. The Corporation mitigates this residual through distributed, multi-jurisdiction custody, multiple independent signers per attestation, scheduled attestation windows, and a fail-closed design that halts critical operations rather than proceed on an uncertain monetary state. It does not claim to eliminate the residual, because no cryptographic system can.
11. Regulatory basis.
The Corporation is established and governed by the Digital Dollar and Monetary Innovation Act of 2029. Validator certification and network governance described in Section 5 implement the Act’s Title IV. Intermediary certification and messaging standards described in Section 9 implement Title V. Reserve, segregation, attestation, and verification requirements described in Sections 7 and 8 implement Title VII. Security and operational-resilience obligations implement Title VIII. Where this document and the Act differ in a future revision, the Act and its implementing regulations govern.
Document control.
| Version | Date | Summary |
|---|---|---|
| 1.0 | February 2031 | Initial public release, concurrent with launch. |
| 1.1 | May 2031 | Added Section 4 (transaction lifecycle) and a threat matrix to Section 10; expanded Sections 5 through 9 with consensus, bridge, attestation-interval, oracle, and intermediary-tier detail. |
Standards and references.
- Board of Governors of the Federal Reserve System. Money and Payments: The U.S. Dollar in the Age of Digital Transformation. 2022.
- Bank for International Settlements, Committee on Payments and Market Infrastructures. Distributed Ledger Technology in Payment, Clearing and Settlement: An Analytical Framework. 2017.
- Castro, M. & Liskov, B. Practical Byzantine Fault Tolerance. Proceedings of OSDI, 1999.
- Laurie, Langley & Kasper. Certificate Transparency. RFC 6962, Internet Engineering Task Force, 2013.
- International Organization for Standardization. ISO 20022 — Universal Financial Industry Message Scheme.
- National Institute of Standards and Technology. Secure Hash Standard (SHS). FIPS PUB 180-4, 2015.
- National Institute of Standards and Technology. Digital Signature Standard (DSS). FIPS PUB 186-4, 2020.
About this document. The USD$ System Architecture overview is a fictional technical publication from the novel The Dearth. The Federal Digital Dollar Corporation, USD$, and the architecture described do not exist.
Generated by Claude, an AI assistant, through the detailed direction of the author.