Skip to content


This page describes the typical actors involved in OAuth 2.0 grants and OpenID Connect flows.

Identity Provider

An identity provider (IdP) is a common service that NAV delegates to for authentication of end-users and/or services. You might come across other terms such as authorization server (AS) or OpenID provider (OP). For simplicity's sake, we'll just use the umbrella term identity provider.

The following is a list of the providers that the platform supports provisioning of clients for:

OpenID Connect:

OAuth 2.0:

Well-Known URL / Metadata Document

This is also referred to as the Discovery Endpoint or Metadata Document Endpoint or other variations with similar names.

Every identity provider will have a metadata document that allows clients to discover the provider's capabilities and endpoints. These are defined in both OpenID Connect Discovery and RFC 8414 - OAuth 2.0 Authorization Server Metadata.

The well-known URL points to this document, which has a registered suffix under the /.well-known path.

An identity provider that supports both OpenID Connect and OAuth 2.0 (such as Azure AD) will usually have the URI suffix /openid-configuration:


A provider that only supports OAuth 2.0 (such as TokenX or Maskinporten) will use the /oauth-authorization-server suffix:


The platform provides these URLs for you at runtime. Most libraries and frameworks support auto-discovery of other relevant endpoints based on the properties this document. We'll take a look at the most essential properties below.

Metadata Document

A metadata document can look like the one shown below. Not all fields are present or required depending on whether the provider supports the OAuth 2.0 or the OpenID Connect specifications.

  "issuer": "",
  "authorization_endpoint": "",
  "token_endpoint": "",
  "end_session_endpoint": "",
  "jwks_uri": "",
  "response_types_supported": [
    "id_token token",
  "response_modes_supported": [
  "id_token_signing_alg_values_supported": [
  "code_challenge_methods_supported": [
  "token_endpoint_auth_methods_supported": [


The issuer property defines the identifier for the provider. This identifier must be a URL that uses the "https" scheme and without query or fragment components.

Any JWT issued by the provider must contain a iss claim with a value that is exactly equal to the value found in the issuer property of the metadata document. This should be validated by OpenID Connect clients and resource servers.

Additionally, the issuer value should be equal to the well-known URL when appending the appropriate well-known suffix. For the examples above, the issuer would be

Token Endpoint

The token_endpoint property points to the endpoint where your client can request tokens from the provider (RFC 6749, Section 3.2). The provider will issue and respond with tokens as defined in RFC 6749, Section 5.1.

Note that the tokens usually have an expiry time (often indicated by the expires_in field in the token response) and are thus suitable for caching if network latency or throughput is a concern.

JWKS Endpoint (Public Keys)

The jwks_uri property points to the endpoint where the provider's public JWKs are published. These public keys are used by your client in order to validate the signature for any JWT issued by the provider.

Do note that the set of keys may change (e.g. revoking/rotating or introducing new keys) at any time. Usually, most libraries are able to handle such cases by regularly fetching a new set of keys from the endpoint.

In cases where a client encounters tokens that are signed with a new key, one will usually just force a refresh of the locally cached set of keys to find the new, matching key. See also these discussions:

Resource Server

A resource server is any entity that requires requests to be authenticated in order to respond with meaningful data. In other words, this is your standard bread-and-butter API server.

Resource servers should require that requests to sensitive endpoints are authenticated with a Bearer token. The server should validate tokens for such requests before accepting or rejecting the request.


A client or application is any entity or device that needs to obtain a token to access a resource server. Tokens are acquired when the client performs authorization grants (OAuth 2) or authentication flows (OpenID Connect).

There are two types of clients that these specifications define; public and confidential. The primary difference here is the notion of whether the client is capable of keeping secrets.

A public client cannot use secrets due to the nature of their runtime environment, such as directly in a browser (in a single-page application) or on a mobile device.

A confidential client can authenticate with an identity provider using a secret or private key due to the secrets being difficult to access to other entities than the client itself. Backend APIs, backend-for-frontends (BFFs) or standalone daemons are typical examples of confidential clients.

Unless specified otherwise, all clients we use are confidential clients.

Client ID

A client ID is a unique identifier associated with your client for a given identity provider. The value of the identifier is generally not considered to be confidential.

Client Authentication

A confidential client must authenticate itself to the identity provider in order to perform grants or flows. There are multiple client authentication methods, however we will only cover the ones used for our identity providers.

Client Secret

A client secret is a password that belongs to a given client.

This is used to authenticate the client when attempting to acquire tokens from the identity provider.

Azure AD is the only identity provider we use that supports the client_secret_post authentication method, which is a simple way of authenticating a client. However, the method bears the risk of exposure and interception as the secret itself is passed in plain-text as part of the request body:

POST /token HTTP/1.1
Content-Type: application/x-www-form-urlencoded


As the platform rotates these credentials relatively regularly, using the client_secret_post method shouldn't pose as a significant security risk. However, we generally do recommend that client assertions are used instead as they're based on public-key cryptography.

Client Assertion

A client assertion is a JWT that can be used to authenticate a client as defined in RFC 7523. Your client must create the assertion and sign it using your client's own private key. The JWT assertion will look something like this (in its decoded form):


  "alg": "RS256",
  "typ": "JWT",
  "kid": "<key identifier for private key>"


  "aud": "",
  "iss": "<your-client-id>",
  "jti": "<some-uuid>",
  "iat": 1520589808,
  "nbf": 1520589808,
  "exp": 1520589928,

This method of authentication (also known as the private_key_jwt method) is supported by all identity providers we use. It is also mandatory for all our clients across all the providers, except Azure AD.

An assertion has several security advantages over a client secret:

  • The client's private key is never exposed or sent as part of a request
  • That the assertion itself is usually only valid for a short duration - meaning that the blast radius is limited if the assertion is intercepted or stolen during transport
  • The provider only needs a public key (which the platform takes care of generating and registering) in order to verify the assertion when receiving an authenticated request from the client.

For example, for the client credentials grant:

POST /token HTTP/1.1
Content-Type: application/x-www-form-urlencoded

eyJpc3Mi[...omitted for brevity...].
J9l-ZhwP[...omitted for brevity...]

Each identity provider may have different requirements for the grant_type and parameter names for the assertions (e.g. assertion vs client_assertion), so make sure to consult the documentation for the specific provider you're using.