Part 2 of this 2-part series explains the did:pkh/CACAO variation for Verite data models and flows, which provides an entry path for wallets that may not support sufficient functionality for emerging decentralized identity patterns

Since some wallets may not themselves be willing to embed protocol-specific logic (interaction with verifiers) or more general verifiable-credential logic, we have to find a kind of "minimum viable" level of support for today’s non-DID-enabled crypto wallets. While handling DIDs and signing Verifiable Presentations brings a kind of secure indirection that enables portability and multi-chain wallets, these properties are not strictly essential to the core functionality of Verite. For this reason, we define a crypto wallet that can receive and pass to dApps a Verifiable Credential issued against its blockchain address adequate, with a few adjustments and supplements.

Phase 0: Issuance Directly to Crypto Wallet​

In a crypto-wallet centric end-to-end flow, the trust model is different and the interplay between credential wallet and crypto wallet can be greatly simplified. The credentials themselves must also be slightly different– instead of obtaining the credential subject DID directly from the wallet to which they are being issued, the issuer will use a credential subject identifier based on a specific blockchain address controlled by that wallet. Using DID terminology, rather than attest to the controller of a wallet, it attests only to a specific address controlled by that wallet.

This greatly simplifies the ownership question, by relying on native mechanisms for proving ownership of the address– at the time of issuance, as well as at time of verification of the credentials.

Two Options of Expressing a Blockchain address as a DID (and as a VC subject)​

Instead of defining the subject of the VC as a chain-agnostic DID provided by the wallet, the issuer will deterministically generate a DID from the blockchain address controlled by the connected wallet. Multiple DID methods allow this possibility; we’ll describe two of them, assuming a wallet with an Ethereum address (referred to as ETH_ADDRESS).

• did:key method - issue against a crypto wallet’s public key: If the issuer has the wallet address ETH_ADDRESS and any signature over a known message, the corresponding public key can be recovered using the ecrecover mechanism ported over from Bitcoin in the early days of Ethereum. In this way. the issuer can deterministically obtain a did:key DID that will be used as the value of credentialSubject.id. This is the method Circle will begin with, for ease of implementation by participants.
  • In this case, the mapping is: did:key:<ecrecover(ETH_ADDRESS, signature)>
  • For blockchains that do not use a Bitcoin-style pay2hash address system, like Solana and Polkadot, no recovery from a signature is necessary because the base form of the address is already a public key supported by multibase and thus by did:key.
• did:pkh method - issue against a crypto wallet’s public address: Other DID methods, like [did:pkh](https://github.com/w3c-ccg/did-pkh/blob/main/did-pkh-method-draft.md), allow DIDs to be defined directly based on blockchain addresses in their commonly-used, public-facing forms. Long term, this is the preferred method. Among other advantages, the implementation across chains is more apparent. - In this case, the mapping is: did:pkh:eip155:1:<ETH_ADDRESS>. eip155 here refers to the EVM namespace (defined in EIP155), while 1 refers to the ethereum mainnet according to EIP155. - Just as the did:key URI scheme relies on the [multibase](https://datatracker.ietf.org/doc/html/draft-multiformats-multibase) registry, so does the did:pkh URI scheme rely on the ChainAgnostic Standards Alliance’s [namespace registry](https://github.com/ChainAgnostic/namespaces) to add non-EVM namespaces. - In cases where human-readability is important, end-users can introspect the VC and see their familiar address, as opposed to a public key that in pay2hash systems like BTC or ETH, they might never have seen or know they control

Sign-In With Ethereum and dApp-native Identity​

Wallets that have not incorporated decentralized identity capabilities rarely support JWT _signing _features, or other token mechanics that are common to the traditional web2 identity world. For this reason, many web3 sites and dApps have started using the wallet connection pattern to create a more feature-rich and auditable session mechanism via offchain message signing (i.e. personal_sign on EVM systems). Equivalents for other blockchain ecosystems, such as Solana, are forthcoming.

In the case of issuance, this signature is enough to extract the crypto wallet’s public key, as mentioned above. Importantly, though, it also enables delegated keys to sign offchain events without another onerous or fraught wallet-interaction, as we will see below.

Phase 1: Off-chain Verification​

Variant: Crypto-Wallet only with only VC storage capabilities​

At verification time, when a wallet "connects” to a dApp by providing an off-chain signature over a structured authentication message, the dApp will have the wallet’s address (and live proof-of-control, if the authentication message included a secure nonce) so it can simply compare this address with the corresponding DID:PKH credentialSubject.id of the VC. This way, the verifier will not need to do an ownership check on the VC, and the dApp can trust the verifier to have received credentials from the right wallet because it, too, will require a wallet connection and prove ownership of the same wallet.

Without necessarily even having to parse, validity-check, or introspect the verifiable credentials, any wallet that can store them (whether as JWTs or even as raw text files) can submit them directly to verifiers, as shown below.

Note: while it is recommended that crypto wallets parse verifiable credentials and check their validity or safety, crypto wallets could in theory allow blind passthrough if the user can assume the responsibility for their contents. In the Verite case, there are little security concerns or abuses possible.

By itself, however, this bare VC is inferior to a VP from a full-featured decentralized-identity wallet, since it does not contain a non-repudiable off-chain wallet signature for auditing purposes. Or, to put it another way, it is only as trustworthy as the authentication of the wallet that sent it to you, and there is little standardization of the receipts you keep of crypto-wallet authentication to replay it for a future auditor or security review.

While the corner cases of impersonation or exfiltrated VCs might be vanishingly rare, the "audit trail” of a bare VC is weaker than a VC wrapped in a timestamped signature. For this reason, we encourage Verite dApps to create a functional equivalent to a verifiable presentation in the form of a signed presentation receipt (signed with a session-specific ephemeral key) for logging purposes. To accomplish this, we return to the Sign-In With Ethereum pattern to elaborate on its key delegation feature.

Sign-In With Ethereum Flow​

As mentioned above, we support the emerging standard approach of the "Sign-In With Ethereum” mechanism which defines a sign-in message that includes a domain-binding anchor, an ephemeral session key, and other caveats. While the ephemeral session key property was not essential to the issuance wallet connection, it can be useful in the verification situation for more trustless (and auditable) logging.

By generating an ephemeral key and including it in the initial wallet-connection message for the crypto wallet to sign over upon authenticating itself to the dApp, the wallet effectively "delegates” that ephemeral key for the duration of the session. In UX terms, this saves the user from a distinct off-chain wallet signature at each confirmation or consent event. Carefully defining the other properties of the SIWE message, dApps can secure and constrain that delegation, link to an applicable terms-of-service agreement, enable DNS-based domain-checks analogous to the "lock symbol” in modern browsers, etc.

Once the user has "connected” their wallet by signing this SIWE message, a CACAO is generated as a receipt of that session (and of the delegation of signing rights to the key). This allows the dApp to use smoother UX than requiring a full off-chain wallet signature to confirm each consent event or internal transaction (such as the presentation of VCs in a signed VP). But it also provides a compact and tamperproof way of encapsulating each event or internal transaction as a time-stamped and signed object for logging purposes– this makes each event as verifiable as an off-chain (or on-chain) signature, via the indirection of the delegated key.

Ownership Verification​

You could say that the crypto wallet delegates the encapsulation and signature of a VP to the dApp, which creates a short-lived key with which to sign the VP, which is a kind of standardized logging object for a presentation event. This allows the verifier to confirm that the dApp is interacting on behalf of the wallet. Since the Verifier has confirmed control of the wallet address with a SIWE message, and the VC is issued to the address directly, there is no ownership verification needed as with a decentralized wallet; thanks to the CACAO, future auditors can replay these interactions and re-check signatures and tamper-proof objects to confirm all of these transactions trustlessly.

Detailed Flow​

1. Wallet initiates DeFi transaction with dApp.
2. dApp generates a session-specific ephemeral signing key and encoded it in the SIWE message for the wallet to sign. This generated session key will delegate the wallet for future signings, once after wallet vouches it (by signing the CACAO).
3. Once the wallet has signed it and returned it to the dApp, the signature and message are encoded into a compact CACAO session receipt for logging and forensic purposes (if needed).
4. Next the dApp lets the verifier know about the session, by POSTing the receipt to an endpoint (eg. signIn). The signed receipt also includes caveats, a domain-binding, an expiration, and other constraints to secure the delegation and the transfer of the session parameters.
5. The verifier saves the CACAO. The verifier only uses this CACAO in the scope of this verification session (to prove the VP signed by the ephemeral key). Once the CACAO verification step is completed, the session object will be updated.
6. Instead of sending the wallet to verify directly with the verifier (as in the previous post), the wallet will submit the VC directly to the dapp (or an agent/service it trusts). The dApp presents the prompt to verify.
7. Wallet submits the bare VC.
8. Subsequent requests from the dApp will include a reference to the session which the verifier can use if they need to check signatures made by that key. The VC(s) submitted by the dApp in this case will not be signed in a VP with the wallet’s key; instead, it/they will be put into a VP and signed by the dApp using the ephemeral key (the signing key mentioned in the first step above) delegated to it by the SIWE message. Introspection into the CACAO is required to extract the public key that signed the VP, as well as a signature from the wallet key over the public key.
9. When all the information is submitted to the verifier, the verifier needs to examine the ownership of the credential:
   1. Extract the public key of the session signing key from the resources section of the CACAO
   2. Use the public key of the session signing key to validate the VP’s signature. This is to ensure that the dApp properly (which held the key and got user consent to delegate signing rights to it) signed the VP and that it has not been tampered with in transport.
   3. Compare iss in CACAO with the wallet’s DID in VC (in this case a did:pkh representing the wallet address as a long URI). They should match, if the dApp’s SIWE message. conforms to the SIWE specification. This is to check the wallet which vouched the session key is the subject (and holder) of the VC, which is also connected to the dApp with a signature over a nonce, included in the CACAO to keep future auditors from having to trust the verifier.

Conclusion:​

Circle’s implementation of the Verite protocol allows us to serve our customers and the dApps they interact with equally, putting the rigor of our KYC processes at the service of a process that is auditable and verifiable end-to-end without duplicating KYC process or PII from those processes across the chain of asset custody. We are proud to be driving the Verite process, and welcome more implementations, whether end-to-end issuer+verifier solutions like ours or more focused implementations that bring more wallets and more users into the ecosystem.

As the Centre team updates its documentation and sample implementation to reflect the new patterns and flows, we will continue to work with them to share the insights we are gaining from our exploratory work with dApps and clients.

Since Verite’s original release, we’ve gotten feedback from development partners and the broader community about which patterns are useful and where guidance could be improved. Some themes include:

• Persistence of verification results to on-chain registries is over-emphasized.
• Verite’s current verification flow documentation assumes that Verifiable Credentials (VCs) are stored in wallets that support decentralized identity protocols, which are currently somewhat scarce in the market.

Upcoming Verite releases (including code and documentation) will address these concerns as follows:

1. A forthcoming editorial revision of Verite documentation will explicitly describe a variant of the subject-submission patterns without any on-chain persistence.
2. A coming release will include code and documentation demonstrating how more of today’s wallets can participate in VC flows in a standardized way, until DIDs become more common.

In this two-part blog series, we’ll preview these updates. This post (Part 1) includes an overview of options for the use of off-chain VCs to support on-chain decision-making, including subject-submission without on-chain persistence. This begins with a summary of core concepts and principles of Verite. Readers already familiar can skip ahead to the "Two Phases of Verification" section for an overview of the new flow being added to the Verite docs and pioneered by Circle’s implementation. After this new end-to-end pattern for Verite is outlined, a more analytical "Discussion" section follows.

The second post in the series will zoom in on how wallets that do not currently support sufficient functionality for emerging decentralized identity patterns can be "retrofitted” and supplemented to support Verite’s off-chain verifications like those described in this post.

Verification Concepts

If you’re new to Verite, or want a reminder, this section describes how Verite credential verification works off-chain to support smart contracts with maximum privacy around sensitive identifying information.

Roles​

The following entities are involved in the Verite credential verification process:

• Crypto wallet: aka "payment wallet”, controls a blockchain address; may be browser or mobile, hosted or self-hosted.
• Credential wallet: aka "identity wallet”, stores and shares VCs, controls a "DID” (meta-wallet identifier). Some or all of these functional roles may be subsumed into a crypto wallet or a trusted dapp, or this entity may be completely distinct software, depending on flow and trust model.
• Verifier: aka the "verifier service” or "verifier endpoint” that consumes a privacy-preserving, off-chain token (a verifiable credential) on behalf of a dApp. It can be operated by the dApp or a trusted partner but in either case it needs to be previously known to and trusted by the dApp at a static address.
• dApp: decides which verifiers to trust; the dApp frontend is responsible for triggering the verification process; the dApp backend is responsible for on-chain validation
• On-chain verification registry (optional): depending on the implementation choices, dApps may rely on on-chain storage of verification results. This may be part of the dApp or a separate registry accessible by other parties. For simplicity, we’ll assume the former case in this blog.

Verification Result​

A Verification Result enables a "verified” crypto wallet – that is, a wallet whose controller has proven possession of a valid VC that has passed a dApp-trusted-verifier’s verification process – to interact on-chain.

You can think of a Verification Result as the result of slicing and dicing a VC, potentially obfuscated, with an attached proof from the Verifier. The Verification Result structure in Verite’s reference implementation includes the following fields, returned along with the Verifier’s cryptographic signature:

• subject: identifies the wallet address
• schema: indicates the kind of credential (flexible according to the implementation)
• expiration: a number that is less than or equal to the expiration value in the VC.

The Verification Result may be extended to include additional fields, which could be carried over identically or fuzzed/obfuscated in some way (which is especially relevant if using on-chain storage).

export type VerificationResult = {
  schema: string
  subject: string  // address
  expiration: number
}

Verifier vs Subject Submission​

Verite’s implementation demonstrates two possible methods of submitting Verification Results to smart contracts; or in other words:

• Verifier-submitted: the verifier submits the Verification Result and signature directly to the smart contract immediately after verification.
• Subject-submitted: the verifier returns the Verification Result and signature to the subject (or an entity or code acting on the subject’s behalf) who then submits the signed Verification Result to the smart contract.

Verification Principles

While verifications can be carried out using a variety of methods, all are expected conform to the following principles:

1. Trusted issuer principle: the credential was issued by one of a list of trusted issuers.
2. Credential ownership principle: The crypto wallet that owns (i.e. can prove themselves to be the subject of) the credential is the one making the DeFi transaction request, regardless of whether the credential itself is held and presented by a separate piece of hardware (i.e. whether the credential wallet and the crypto wallet are distinct)
3. Trusted verifier principle: The verifier that provides verification is on the allowlist of verifiers already trusted by the dApp.

The fact that all Verified Credentials are signed by their self-identifying issuers makes the first criteria simple: issuers’ public keys can be obtained from the identifiers listed in a registry of trusted issuers, and their signatures can be checked against them at verification time. In this way, the verifier enforces the trusted issuer principle.

DApps, in turn, uphold the trusted verifier principle, deciding which verifiers they trust to verify VCs – including standard VCs checks (non-revocation, tamper-free) and any other fitness-for-purpose checks (such as subsetting valid Verite credentials according to additional constraints on issuers or KYC processes).

Enforcement of the credential ownership principle will vary depending on how the crypto wallet relates to the credential wallet. In this post, we’ll describe how this works with a distinct decentralized identity wallet controlled by the same actor as a crypto wallet; in the next, we’ll describe how crypto-wallets-only actors may be accommodated.

Two Phases of Verification

Verite’s Smart Contract Patterns documentation describes the two-phase process of how smart contracts can require successful verification of VCs. To summarize:

• Phase 1: "Off-chain” verification by a verifier, based on traditional web or API stacks, resulting in a lightweight "Verification Result” object.
• Phase 2: On-chain validation of the Verification Result, which is optionally persisted on-chain in private, obfuscated, or opaque form referred to as a "Verification Record”.

The two-phase process enables the use of VCs with chain-based applications for which verification operations are not technically possible and/or economically practical. More importantly, this reduces (by construction) the amount of potentially sensitive data sent on-chain. All examples that follow will refer to main-net Ethereum, but all these patterns should be replicable on all major chains, availing themselves of further privacy or obfuscation techniques where supported.

The subsequent sections discuss these phases and variations. This first post in the two-part series assumes a decentralized identity wallet, and the next will cover other options.

Phase 1: Off-chain Verification

Assuming a Decentralized Identity+Crypto Wallet

1. The wallet initiates a DeFi transaction with a dApp.

2. The dApp (frontend) chooses a trusted verifier to start the verification process.

3. The verifier responds to the wallet, telling it how to start the process¹ through the dApp.

4. The wallet chooses one of the credentials stored locally and submits it to the verifier.

5. The verifier confirmed the credential’s signature and the validity of its contents. The dApp retrieves (via callback or polling) confirmation.

6. DApp unlocks the subsequent DeFi transaction flow.

Ownership verification​

In step 4, when the wallet submits the credential, the wallet wraps the VC with a Verifiable Presentation (VP) and signs it. When the verifier receives the credential, it uses the Decentralized Identifier (DID) – for example, did:key:<public_key>² – listed as holder in the VP to verify the JWT’s signature (made from the corresponding private key). It also compares this DID in the VP with the DID listed as credentialSubject.id in the VC, i.e. the public key that the credential was issued against. Those two checks guarantee the wallet, submitting the credentials, was the same wallet that the issuer intended, and thus rightly owns it.

Phase 2: Validation and use in smart contracts

The second phase of verification covers how a Verification Result gets used as a prerequisite to on-chain interaction.

Validating a Verification Result​

Along with a verification result, a smart contract will receive a signature on the result, allowing them to confirm that the result was provided by an authorized party, as described here.

On-chain storage of Verification Results (or not)​

We’ll look at 3 cases below. With the verifier-submitted pattern, the verifier always stores the Verification Result on-chain. With subject-submission, the Verification Result may or may not be stored on-chain, depending on the implementation.

Case 1: Verifier-submitted verification result, stored on-chain​

With the verifier-submitted approach, the verifier stores the Verification Result on-chain after verification.

We’ll cover step 1.5 later; this is relevant for subsequent interactions with a dApp when on-chain storage is used.

Case 2: Subject-submitted verification result, stored on-chain​

The subject-submitted approach also supports persistence, as determined by the implementor. Like the verifier-submitted case, in subsequent interactions, the dApp will be able to determine that the wallet is already registered.

We’ll cover step 1.5 later; this is relevant for subsequent interactions with a dApp when on-chain storage is used.

Case 3: Subject-submitted verification result, off-chain​

If the Verification Result is not persisted on-chain, then every interaction of the wallet with the contract will be associated with a fresh, unique Verification Result.

Subsequent interactions

Assuming the implementation is using on-chain storage, after the phases 1 & 2 (plus on-chain storage) are completed, then on subsequent interactions, a dApp could check the registry to confirm the wallet/address is already approved, avoiding the need to re-verify the VC. The dApp backend (or registry, if separate) can simply check that msg.sender is in the set of registered addresses.

Without on-chain storage, in general the VC would be re-verified³ each time (as shown in the diagram in Case 3), and the verification result submitted along with the transaction. Since verification is off-chain (and generally fast/inexpensive, depending on the provider), and since this avoids on-chain storage of potentially correlatable data, this is often the preferred solution.

Discussion

On-chain Storage Considerations​

The on-chain storage patterns have implications that require careful consideration:

1. On-chain persistence of any potentially identifiable (or re-identifiable) data should be avoided. If on-chain persistence is used, it is up to the design of a particular implementation to manage that smart contract’s access rights and obfuscation mechanisms to minimize privacy risks.
2. The verifier-submitted pattern, in its current form, assumes the verifier’s willingness to pay transaction fees (e.g. "gas fees”, in EVM systems).
3. The verifier-submitted pattern includes an obligation that the verifier update the on-chain state corresponding to revoked credentials.
4. In general, this may not be the most cost-effective option.

There are optimizations – in the Verite open source, the Verification Registry may be used among multiple contracts, and proposed improvements enable further reuse through proxies. These are all possible for implementations; Verite’s open source repositories are intended as example implementations, and not as normative guidance.

However, we think it’s important to draw attention to off-chain options, as it has a profile of privacy characteristics we find favorable in many use-cases. This is one of the reasons we’re increasingly demonstrating this option, as seen in the off-chain NFT allowlist.

Circle’s Implementation Choices​

It was not necessary to write the verification results to on-chain storage in Circle’s implementation. The advantages of flushing to chains include the avoidance of repeated verification for the same wallets. However, the advantages do not always outweigh the drawbacks:

1. Short TTL is required for verification because a user's KYC status could change.
2. Chain writers must pay a gas fee.

As a result, Circle chose and suggests other implementers also choose this approach: do not write verification results into chains unless there is a substantial efficiency gain, i.e., if many transactions will be authorized by the average on-chain write operations.

Since Circle’s architecture is optimized for delivering verification reliably and simply to dApps, we choose to explain the design choices we’ve made from the perspective of the verification process, where valid Verifiable Credentials get communicated to smart-contracts powering dApps.

Up Next​

In the following post, we’ll describe options for wallets that don’t understand Decentralized Identity protocols to participate in VC exchanges, while conforming to the verification principles above.

The Verite core team is grateful to Circle’s Verite engineers Tao Tao, Daniel Lim, and Fei Niu for taking the lead on this exciting work!