Client keys
All client requests in Open Payments are signed using a unique key that identifies the client to the authorization and resource servers. All requests, except for new grant requests, also carry an access token that is bound to the key.
Key registry
A key registry is a list of keys, generated and stored by the client, for when the client requires access to protected Open Payment resources. Since grant requests are completed over multiple signed HTTP requests, it’s important for the client to provide a way to consistently identify itself across these requests to the authorization server. The key registry allows the authorization server to verify that the client is who it says it is.
Each client is represented by a wallet address. A client’s key registry is publicly accessible through its wallet address via a jwks.json
endpoint. An authorization server can retrieve the client’s key registry by accessing WALLET_ADDRESS/jwks.json
.
Example
https://wallet.example.com/alice/jwks.json
Registry structure
The key registry must expose public keys in the form of JSON Web Key Sets (JWKS). The keys must be generated using the Ed25519
algorithm and the resultant JWKS document must contain the following fields and values.
Additionally, it must contain the x
and kid
(key ID) fields for the specific client to identify itself in a signature.
Example: https://wallet.example.com/alice/jwks.json
Key generation
Before a client can request a grant for the first time, it must:
- Generate an asymmetric key pair. Keys must be generated using the ed25519 algorithm.
- Add the public key to its key registry.
- Store the private key. The private key is used to sign the payload described in the key proofing method section below.
Client requests
Since client requests are completed over multiple signed HTTP requests, it’s important for a client to provide a way to consistently identify itself across these requests. As such, clients must include the following when making requests:
- Headers
- A
Signature-Input
header that includes thekeyId
associated with the client’s key pair. This header is a comma-separated list of headers that map to values in the data that was signed. - A
signature
header generated based on theSignature-Input
, using theEdDSA
signing algorithm
- A
- Body
- A
client
property containing the client’s wallet address
- A
Securing client requests follows a profile of what’s defined in the GNAP specification.
Grant requests
Upon receiving a signed grant request, the authorization server obtains the client’s domain from the client
property. The authorization server binds the domain to the grant in order to use the domain to acquire the key set for subsequent grant requests.
The authorization server then acquires the client’s key registry by making a GET
request to the client’s JWKS endpoint at WALLET_ADDRESS/jwks.json
. When the authorization server locates the public key containing the keyId
included in the request’s Signature-Input
header, the authorization server uses the key to decrypt and validate the request’s signature. This binds the client to the grant and allows the authorization server to continue with the grant request.
Key proofing method
HTTP message signatures
Open Payments uses the HTTP message signatures (httpsig
) key proofing method.
The httpsig
proofing method must be declared as part of the key material when directly using a key to request a grant. The key material below is for illustrative purposes. In Open Payments, it’s expected that the wallet address be used in the grant request.
Example
When using httpsig
, the signer (the client) creates an HTTP message signature. Open Payments clients typically secure their requests to servers by presenting an access token and proof of a key it possesses. The exception is for calls to an authorization server to initiate a grant. In this case, a key proof is used with no access token and is a non-authorized signed request.
See the HTTP message signatures page for more information specific to Open Payments. Additional information is found in the specification for HTTP message signatures.
Sequence diagram
sequenceDiagram Client->>Client: Generate keys, add public key to registry Client->>Authorization server: Make initial grant request (POST /)
and sign request with private key Authorization server->>Authorization server: Pull the keyId from the grant request's signature-input header,
get client domain from initial grant request's body Authorization server->>Client: GET {client_domain/jwks.json} Client->>Authorization server: Return registry's public key Authorization server->>Authorization server: Find the registry key with the matching keyId,
validate signature with key, then bind client to grant Authorization server->>Client: Complete initial grant request Client->>Authorization server: Make grant continuation request (POST /continue),
sign with private key Authorization server->>Authorization server: Pull the keyId from the grant request's signature-input header,
get client domain from the database entry for the grant Authorization server->>Client: GET {client_domain/jwks.json} Client->>Authorization server: Return registry's public key Authorization server->>Authorization server: Find the registry key with the matching keyId,
validate signature with key Authorization server->>Client: Complete grant continuation request