OpenID Connect Authenticator

The OpenID Connect (OIDC) authenticator allows users to be authenticated by any external OIDC Provider (OP) that supports the following aspects of the standard:

  • OpenID Connect Discovery 1.0 whereby the OP publishes its configuration
  • The authorization code flow
  • Signed ID Tokens
  • Encrypted ID Tokens
  • Fetching additional claims from Userinfo endpoint of OP
  • JSON Web Key Set published at a jwks_uri

A client needs to be registered at the OP, and have a redirect URI configured as described in the next section.

The Redirect URI

The redirect URI is the URI where the OP will redirect the user to when the authentication is done. The URI that the Authentication Service uses is generated automatically based on configuration, and is the path to the authenticator with an added /callback. This last part of the path is only used in the communication between the Authentication Service and the OP, and can not be changed whereas the preceding components of the path can be.

The form is like this:

<base-url>/<authentication-endpoint>/<authenticator-id>/callback

For example, for a server running on server1.local and a protocol of https and listening on port 8443 and having an authentication endpoint configured to be /authn/authenticate and with an OIDC authenticator instance called oidc1, the redirect URI would be:

https://server1.local:8443/authn/authenticate/oidc1/callback

Redirect URI Override

The redirect URI that will be used in authorization (and token) requests can be overridden by setting the redirect-uri-override configuration value. Note that the real redirect URI of the OIDC authenticator will not be changed. This means that there has to be a component that relays the authorization response to the real redirect URI of the OIDC authenticator. Overriding the redirect URI may be useful when migrating to the OIDC authenticator so that existing clients don’t have to change the redirect URI immediately.

JWKS Endpoint

The OIDC authenticator publishes its JSON Web Key Set on the anonymous endpoint of the authenticator, under the /jwks subpath. For example, if an authenticator with ID my-oidc-authenticator is configured on an authentication profile that has the anonymous endpoint mapped on /authenticate/anonymous, then the metadata is published on the path /authenticate/anonymous/my-oidc-authenticator/jwks.

When used as a dynamic authenticator the metadata endpoint expects the federation id to be passed in the query parameter (fid). When the federation id is not provided the default value is used. For example, the request for JWKS with federation ID test-federation-id and dynamic authenticator ID my-oidc-dynamic-authenticator, (configured on an authentication profile that has the anonymous endpoint mapped on /authenticate/anonymous), will be the following /authenticate/anonymous/my-oidc-dynamic-authenticator/jwks?fid=test-federation-id.

Returned attributes

The claims from the returned ID Token describing the user will be returned in the Subject Attributes, as well as the attributes returned from the Userinfo endpoint if configured. Additional attributes will be added to the context attributes, described in the table below. All values are optional and depends on if the OP returns them.

Context Attribute Name OP Attribute Name Value From
iss iss ID Token
aud aud ID Token
exp exp ID Token
iat iat ID Token
auth_time auth_time ID Token
amr amr ID Token
op_acr acr ID Token
op_access_token access_token Token Response
op_refresh_token refresh_token Token Response

Parameter Mappings

Since version 9.4, the oidc authenticator supports adding custom query parameters in the authorization request. Those parameters can take their value either from a query parameter of the original request to the Curity Identity Server or have a static value from configuration.

Note

If use-value-from-request is used, the authorization request to the OpenID Provider will include the custom query parameter only if the original request to the Curity Identity Server contains the configured query parameter.

Configuration

The general authenticator configuration is described under Authenticator Base Configuration. This section describes the additional parameters that are available for the OIDC authenticator.

Important

Authenticator base is /profiles/profile{id type}/settings/authentication-service/authenticators/

List Entry: <authenticator_base>/authenticator/ (key id)

configuration-url: The url to the openid-configuration document at the OP

type string

mandatory

client_id: The OAuth 2 client ID that is registered at the OP

type string

mandatory

use-http-basic-authentication: When enabled, the client_id and client_secret are sent in a RFC2617 compliant header instead of the request-body

type boolean

enabled only when client_secret is used as a Client Authentication Method

Client authentication method (a choice of the following):

mandatory

client_secret: The OAuth 2 client secret that is registered at the OP
type string
asymmetrically-signed-jwt: The asymmetrically signed JWT (private_key_jwt) settings
signing-key: Signing key for the asymmetrically signed JWT (private_key_jwt)

mandatory true

type leafref: "/base:facilities/base:crypto/base:signing-keys/base:signing-key/base:id"

signature-algorithm: Signature algorithm for the asymmetrically signed JWT (private_key_jwt)

mandatory true

type enum

symmetrically-signed-jwt: The symmetrically signed JWT (client_secret_jwt) settings
signing-key: Signing key for the symmetrically signed JWT (client_secret_jwt)

mandatory true

type leafref: "/base:facilities/base:crypto/base:signing-keys/base:signing-key/base:id"

signature-algorithm: The signature algorithms to allow for JWT (client_secret_jwt)

mandatory true

type enum

Note: Both asymmetrically and symmetrically signed JWT are sent to the OP endpoint as the client_assertion parameter.

scope: The scopes to ask the OP for as a space-separated list

type string

default openid

clock_skew: The allowed clock-skew in seconds when validating the JWT from the OP

type uint32

default 60

authentication-context-class-reference: The space-separated list of authentication methods that the OP should use
type string
redirect-uri-override: An optional override of the redirect URI that will be used in the authorization requests. The OP will redirect to this URI, however the actual redirect URI of this authenticator will remain unchanged. This means that the authorization response has to be relayed to the actual redirect URI of this authenticator.
type inet:uri

Example (asymmetrically-signed-jwt):

<authenticator xmlns="https://curity.se/ns/conf/profile/authentication">
    <id>oidc</id>
    <oidc xmlns="https://curity.se/ns/conf/authenticators/oidc">
        <configuration-url>https://example.com/.well-known/openid-configuration</configuration-url>
        <client-id>MyOidcClient</client-id>
        <asymmetrically-signed-jwt>
            <signing-key>oidc-signing-key</signing-key>
            <signature-algorithm>RS256</signature-algorithm>
        </asymmetrically-signed-jwt>
        <authentication-context-class-reference>example-acr</authentication-context-class-reference>
    </oidc>
</authenticator>

Example (symetrically-signed-jwt):

<authenticator xmlns="https://curity.se/ns/conf/profile/authentication">
    <id>oidc</id>
    <oidc xmlns="https://curity.se/ns/conf/authenticators/oidc">
        <configuration-url>https://example.com/.well-known/openid-configuration</configuration-url>
        <client-id>MyOidcClient</client-id>
         <symmetrically-signed-jwt>
            <signing-key>oidc-signing-key</signing-key>
            <signature-algorithm>HS256</signature-algorithm>
        </symmetrically-signed-jwt>
        <authentication-context-class-reference>example-acr</authentication-context-class-reference>
    </oidc>
</authenticator>

Example (secret):

<authenticator xmlns="https://curity.se/ns/conf/profile/authentication">
    <id>oidc</id>
    <oidc xmlns="https://curity.se/ns/conf/authenticators/oidc">
        <configuration-url>https://example.com/.well-known/openid-configuration</configuration-url>
        <client-secret>123456778912345677890012345677890</client-secret>
        <client-id>MyOidcClient</client-id>
        <authentication-context-class-reference>example-acr</authentication-context-class-reference>
    </oidc>
</authenticator>

encrypted-id-token Some providers return encrypted ID-tokens (JWE). To handle those, a decryption key needs to be configured. Additionally, two whitelists can be added.

allowed-algorithms: A whitelist for the key management algorithms. These algorithms are used to derive the actual content encryption key. Empty list means any algorithm can be used. type list

allowed-content-encryption-algorithms: A whitelist for the allowed decryption algorithms. Any of these algorithms can be used to decrypt the token. Empty list means any algorithm can be used. type list

fetch-userinfo The authenticator can be configured to fetch additional claims from the Userinfo endpoint of the OpenID provider. The format of the response can be different depending on the provider, and the expected format needs to be configured for this authenticator. The following formats are supported:

plain: The response is in plain JSON.

signed: The response is a signed JWT (JWS). The signature and expiry of the JWT will be validated. The public key will be collected from the metadata.

encrypted: The response is an encrypted JWT (JWE). The JWT will be decrypted using the configured key, and the expiry will be validated.

Decryption key is mandatory

The payload of the decrypted JWT can be be of two formats.

  • Signed JWT. The signature and expiry of the JWT will be validated. They public key will be collected from the metadata.
  • Plain JSON.

For encrypted userinfo, it’s also possible to configure allowed-content-encryption-algorithms and allowed-algorithms

Important

If the response is of any other format than the configured one, the authentication will abort.

When a signed JWT is used in the UserInfo response, it must, by default, include the issuer (iss) and audience (aud) claims. If a misbehaving OpenID provider is used, it’s possible to allow those claims to be missing by setting strict-validation-when-signed to false. It’s recommended to not use this setting and try to correct the provider.

The result of the authentication will be the Userinfo claims, merged with the claims from the ID token. When a claim with the same name exists on both sets of claims, the resulting value is determined as follows:

  • When both source values are JSON maps/objects - the map keys/values are merged recursively, as described here, and the resulting map is used as the claim value.
  • When both source values are JSON arrays - the claim value is the concatenation of the two arrays.
  • Otherwise - the claim value from Userinfo is used.
parameter-mappings:

type list

parameter-mapping: <name> The name of the parameter mapping

type string

static-value: A static string to use as the value.
type string
use-value-from-request: The name of the query parameter of the original request to get the value from.
type string

Example (parameter-mappings):

<authenticator>
  <id>oidc</id>
  <oidc xmlns="https://curity.se/ns/conf/authenticators/oidc">
    <parameter-mappings>
      <parameter-mapping>
        <parameter-name>queryParameter1</parameter-name>
        <static-value>abcdefg</static-value>
      </parameter-mapping>
      <parameter-mapping>
        <parameter-name>queryParameter2</parameter-name>
        <use-value-from-request>state</use-value-from-request>
      </parameter-mapping>
    </parameter-mappings>
  </oidc>
</authenticator>