OpenID Connect authorization code flow mechanism for protecting web applications

The OIDC web-app application requires URLs of the OIDC provider’s authorization, token, JsonWebKey (JWK) set, and possibly the UserInfo, introspection and end-session (RP-initiated logout) endpoints.

By convention, they are discovered by adding a /.well-known/openid-configuration path to the configured quarkus.oidc.auth-server-url.

Alternatively, if the discovery endpoint is not available, or you prefer to reduce the discovery endpoint round-trip, you can disable endpoint discovery and configure relative path values. For example:

Some OIDC providers support metadata discovery but do not return all the endpoint URL values required for the authorization code flow to complete or to support application functions, for example, user logout. To work around this limitation, you can configure the missing endpoint URL values locally, as outlined in the following example:

You can use this same configuration to override a discovered endpoint URL if that URL does not work for the local Quarkus endpoint and a more specific value is required. For example, a provider that supports both global and application-specific end-session endpoints returns a global end-session URL such as http://localhost:8180/oidcprovider/account/global-logout. This URL will log the user out of all the applications into which the user is currently logged in. However, if the requirement is for the current application to log the user out of a specific application only, you can override the global end-session URL, by setting the quarkus.oidc.end-session-path=logout parameter.

OIDC providers typically require applications to be identified and authenticated when they interact with the OIDC endpoints. Quarkus OIDC, specifically the quarkus.oidc.runtime.OidcProviderClient class, authenticates to the OIDC provider when the authorization code must be exchanged for the ID, access, and refresh tokens, or when the ID and access tokens must be refreshed or introspected.

Typically, client id and client secrets are defined for a given application when it enlists to the OIDC provider. All OIDC client authentication options are supported. For example:

Or:

The following example shows the secret retrieved from a credentials provider:

Example of private_key_jwt with the PEM key inlined in application.properties, and where the signature algorithm is RS256:

Example of private_key_jwt with the PEM key file, and where the signature algorithm is RS256:

Using client_secret_jwt or private_key_jwt authentication methods ensures that a client secret does not get sent to the OIDC provider, therefore avoiding the risk of a secret being intercepted by a 'man-in-the-middle' attack.

You can filter OIDC requests made by Quarkus to the OIDC provider by registering one or more OidcRequestFilter implementations, which can update or add new request headers and can also log requests.

例如：

Alternatively, you can use an @OidcEndpoint annotation to apply this filter to responses from the OIDC discovery endpoint only:

OidcRequestContextProperties can be used to access request properties. Currently, you can use a tenand_id key to access the OIDC tenant id and a grant_type key to access the grant type which the OIDC provider uses to acquire tokens. The grant_type can only be set to either authorization_code or refresh_token grant type, when requests are made to the token endpoint. It is null in all other cases.

You can filter responses from the OIDC providers by registering one or more OidcResponseFilter implementations, which can check the response status, headers and body in order to log them or perform other actions.

You can have a single filter intercepting all the OIDC responses, or use an @OidcEndpoint annotation to apply this filter to the specific endpoint responses only. For example:

When a user is redirected to the OIDC provider to authenticate, the redirect URL includes a redirect_uri query parameter, which indicates to the provider where the user has to be redirected to when the authentication is complete. In our case, this is the Quarkus application.

Quarkus sets this parameter to the current application request URL by default. For example, if a user is trying to access a Quarkus service endpoint at http://localhost:8080/service/1, then the redirect_uri parameter is set to http://localhost:8080/service/1. Similarly, if the request URL is http://localhost:8080/service/2, then the redirect_uri parameter is set to http://localhost:8080/service/2.

Some OIDC providers require the redirect_uri to have the same value for a given application, for example, http://localhost:8080/service/callback, for all the redirect URLs. In such cases, a quarkus.oidc.authentication.redirect-path property has to be set. For example, quarkus.oidc.authentication.redirect-path=/service/callback, and Quarkus will set the redirect_uri parameter to an absolute URL such as http://localhost:8080/service/callback, which will be the same regardless of the current request URL.

If quarkus.oidc.authentication.redirect-path is set, but you need the original request URL to be restored after the user is redirected back to a unique callback URL, for example, http://localhost:8080/service/callback, set quarkus.oidc.authentication.restore-path-after-redirect property to true. This will restore the request URL such as http://localhost:8080/service/1.

You can register one or more io.quarkus.oidc.OidcRedirectFilter implementations to filter OIDC redirects to OIDC authorization and logout endpoints but also local redirects to custom error and session expired pages. Custom OidcRedirectFilter can add additional query parameters, response headers and set new cookies.

For example, the following simple custom OidcRedirectFilter adds an additional query parameter and a custom response header for all redirect requests that can be done by Quarkus OIDC:

See also the Customizing authentication requests section how to configure additional query parameters for OIDC authorization point.

Custom OidcRedirectFilter for local error and session expired pages can also create secure cookies to help with generating such pages.

For example, let’s assume you need to redirect the current user whose session has expired to a custom session expired page available at http://localhost:8080/session-expired-page. The following custom OidcRedirectFilter encrypts the user name in a custom session_expired cookie using an OIDC tenant client secret:

Next, a public JAX-RS resource which generates session expired pages can use this cookie to create a page tailored for this user and the corresponding OIDC tenant, for example:

You can access information about authorization in different ways.

A core part of the authentication process is ensuring the chain of trust and validity of the information. This is done by ensuring tokens can be trusted.

Proof Key for Code Exchange (PKCE) minimizes the risk of authorization code interception.

While PKCE is of primary importance to public OIDC clients, such as SPA scripts running in a browser, it can also provide extra protection to Quarkus OIDC web-app applications. With PKCE, Quarkus OIDC web-app applications act as confidential OIDC clients that can securely store the client secret and use it to exchange the code for the tokens.

You can enable PKCE for your OIDC web-app endpoint with a quarkus.oidc.authentication.pkce-required property and a 32-character secret that is required to encrypt the PKCE code verifier in the state cookie, as shown in the following example:

If you already have a 32-character client secret, you do not need to set the quarkus.oidc.authentication.pkce-secret property unless you prefer to use a different secret key. This secret will be auto-generated if it is not configured and if the fallback to the client secret is not possible in cases where the client secret is less than 16 characters long.

The secret key is required to encrypt a randomly generated PKCE code_verifier while the user is redirected with the code_challenge query parameter to an OIDC provider to authenticate. The code_verifier is decrypted when the user is redirected back to Quarkus and sent to the token endpoint alongside the code, client secret, and other parameters to complete the code exchange. The provider will fail the code exchange if a SHA256 digest of the code_verifier does not match the code_challenge that was provided during the authentication request.

Another important requirement for authentication is to ensure that the data the session is based on is up-to-date without requiring the user to authenticate for every single request. There are also situations where a logout event is explicitly requested. Use the following key points to find the right balance for securing your Quarkus applications:

There are two main ways for the authentication information to expire: the tokens expired and were not renewed or an explicit logout operation was triggered.

Let’s start with explicit logout operations.

Some well-known providers such as GitHub or LinkedIn are not OpenID Connect providers, but OAuth2 providers that support the authorization code flow. For example, GitHub OAuth2 and LinkedIn OAuth2. Remember, OIDC is built on top of OAuth2.

The main difference between OIDC and OAuth2 providers is that OIDC providers return an ID Token that represents a user authentication, in addition to the standard authorization code flow access and refresh tokens returned by OAuth2 providers.

OAuth2 providers such as GitHub do not return IdToken, and the user authentication is implicit and indirectly represented by the access token. This access token represents an authenticated user authorizing the current Quarkus web-app application to access some data on behalf of the authenticated user.

For OIDC, you validate the ID token as proof of authentication validity whereas in the case of OAuth2, you validate the access token. This is done by subsequently calling an endpoint that requires the access token and that typically returns user information. This approach is similar to the OIDC UserInfo approach, with UserInfo fetched by Quarkus OIDC on your behalf.

For example, when working with GitHub, the Quarkus endpoint can acquire an access token, which allows the Quarkus endpoint to request a GitHub profile for the current user.

To support the integration with such OAuth2 servers, quarkus-oidc needs to be configured a bit differently to allow the authorization code flow responses without IdToken: quarkus.oidc.authentication.id-token-required=false.

The next step is to ensure that the returned access token can be useful and is valid to the current Quarkus endpoint. The first way is to call the OAuth2 provider introspection endpoint by configuring quarkus.oidc.introspection-path, if the provider offers such an endpoint. In this case, you can use the access token as a source of roles using quarkus.oidc.roles.source=accesstoken. If no introspection endpoint is present, you can attempt instead to request UserInfo from the provider as it will at least validate the access token. To do so, specify quarkus.oidc.token.verify-access-token-with-user-info=true. You also need to set the quarkus.oidc.user-info-path property to a URL endpoint that fetches the user info (or to an endpoint protected by the access token). For GitHub, since it does not have an introspection endpoint, requesting the UserInfo is required.

OAuth2 servers might not support a well-known configuration endpoint. In this case, you must disable the discovery and configure the authorization, token, and introspection and UserInfo endpoint paths manually.

For well-known OIDC or OAuth2 providers, such as Apple, Facebook, GitHub, Google, Microsoft, Spotify, and X (formerly Twitter), Quarkus can help significantly simplify your application’s configuration with the quarkus.oidc.provider property. Here is how you can integrate quarkus-oidc with GitHub after you have created a GitHub OAuth application. Configure your Quarkus endpoint like this:

For more information about configuring other well-known providers, see OpenID Connect providers.

This is all that is needed for an endpoint like this one to return the currently-authenticated user’s profile with GET http://localhost:8080/github/userinfo and access it as the individual UserInfo properties:

If you support more than one social provider with the help of OpenID Connect Multi-Tenancy, for example, Google, which is an OIDC provider that returns IdToken, and GitHub, which is an OAuth2 provider that does not return IdToken and only allows access to UserInfo, then you can have your endpoint working with only the injected SecurityIdentity for both Google and GitHub flows. A simple augmentation of SecurityIdentity will be required where a principal created with the internally-generated IdToken will be replaced with the UserInfo-based principal when the GitHub flow is active:

Now, the following code will work when the user signs into your application by using Google or GitHub:

Possibly a simpler alternative is to inject both @IdToken JsonWebToken and UserInfo and use JsonWebToken when handling the providers that return IdToken and use UserInfo with the providers that do not return IdToken.

You must ensure that the callback path you enter in the GitHub OAuth application configuration matches the endpoint path where you want the user to be redirected after a successful GitHub authentication and application authorization. In this case, it has to be set to http://localhost:8080/github/userinfo.

You can register the @ApplicationScoped bean which will observe important OIDC authentication events. When a user logs in for the first time, re-authenticates, or refreshes the session, the listener is updated. In the future, more events might be reported. For example:

For information about Authorization Code Flow access token propagation to downstream services, see the Token Propagation section.

You can check if implementing single-page applications (SPAs) the way it is suggested in the Single-page applications section of the "OpenID Connect (OIDC) Bearer token authentication" guide meets your requirements.

If you prefer to use SPAs and JavaScript APIs such as Fetch or XMLHttpRequest(XHR) with Quarkus web applications, be aware that OIDC providers might not support cross-origin resource sharing (CORS) for authorization endpoints where the users are authenticated after a redirect from Quarkus. This will lead to authentication failures if the Quarkus application and the OIDC provider are hosted on different HTTP domains, ports, or both.

In such cases, set the quarkus.oidc.authentication.java-script-auto-redirect property to false, which will instruct Quarkus to return a 499 status code and a WWW-Authenticate header with the OIDC value.

The browser script must set a header to identify the current request as a JavaScript request for a 499 status code to be returned when the quarkus.oidc.authentication.java-script-auto-redirect property is set to false.

If the script engine sets an engine-specific request header itself, then you can register a custom quarkus.oidc.JavaScriptRequestChecker bean, which will inform Quarkus if the current request is a JavaScript request. For example, if the JavaScript engine sets a header such as HX-Request: true, then you can have it checked like this:

and reload the last requested page in case of a 499 status code.

Otherwise, you must also update the browser script to set the X-Requested-With header with the JavaScript value and reload the last requested page in case of a 499 status code.

例如：

If you plan to consume this application from a single-page application running on a different domain, you need to configure cross-origin resource sharing (CORS). For more information, see the CORS filter section of the "Cross-origin resource sharing" guide.

The OIDC authentication mechanism can be affected if your Quarkus application is running behind a reverse proxy, gateway, or firewall when HTTP Host header might be reset to the internal IP address and HTTPS connection might be terminated, and so on. For example, an authorization code flow redirect_uri parameter might be set to the internal host instead of the expected external one.

In such cases, configuring Quarkus to recognize the original headers forwarded by the proxy will be required. For more information, see the Running behind a reverse proxy Vert.x documentation section.

For example, if your Quarkus endpoint runs in a cluster behind Kubernetes Ingress, then a redirect from the OIDC provider back to this endpoint might not work because the calculated redirect_uri parameter might point to the internal endpoint address. You can resolve this problem by using the following configuration, where X-ORIGINAL-HOST is set by Kubernetes Ingress to represent the external endpoint address.:

quarkus.oidc.authentication.force-redirect-https-scheme property can also be used when the Quarkus application is running behind an SSL terminating reverse proxy.

The OIDC provider externally-accessible authorization, logout, and other endpoints can have different HTTP(S) URLs compared to the URLs auto-discovered or configured relative to the quarkus.oidc.auth-server-url internal URL. In such cases, the endpoint might report an issuer verification failure and redirects to the externally-accessible OIDC provider endpoints might fail.

If you work with Keycloak, then start it with a KEYCLOAK_FRONTEND_URL system property set to the externally-accessible base URL. If you work with other OIDC providers, check the documentation of your provider.

OIDC providers behind a firewall may redirect Quarkus OIDC HTTP client’s GET requests to some of its endpoints such as a well-known configuration endpoint. By default, Quarkus OIDC HTTP client follows HTTP redirects automatically, excluding cookies which may have been set during the redirect request for security reasons.

If you would like, you can disable it with quarkus.oidc.follow-redirects=false.

When following redirects automatically is disabled, and Quarkus OIDC HTTP client receives a redirect request, it will attempt to recover only once by following the redirect URI, but only if it is exactly the same as the original request URI, and as long as one or more cookies were set during the redirect request.

添加以下依赖关系：

Prepare the REST test endpoints and set application.properties. For example:

Finally, write the test code, for example:

OidcWiremockTestResource recognizes alice and admin users. The user alice has the user role only by default - it can be customized with a quarkus.test.oidc.token.user-roles system property. The user admin has the user and admin roles by default - it can be customized with a quarkus.test.oidc.token.admin-roles system property.

Additionally, OidcWiremockTestResource sets the token issuer and audience to https://service.example.com, which can be customized with quarkus.test.oidc.token.issuer and quarkus.test.oidc.token.audience system properties.

OidcWiremockTestResource can be used to emulate all OIDC providers.

Using Dev Services for Keycloak is recommended for integration testing against Keycloak. Dev Services for Keycloak will start and initialize a test container: it will create a quarkus realm, a quarkus-app client (secret secret), and add alice (admin and user roles) and bob (user role) users, where all of these properties can be customized.

First, prepare application.properties. You can start with a completely empty application.properties file as Dev Services for Keycloak will register quarkus.oidc.auth-server-url pointing to the running test container as well as quarkus.oidc.client-id=quarkus-app and quarkus.oidc.credentials.secret=secret.

However, if you already have all the required quarkus-oidc properties configured, then you only need to associate quarkus.oidc.auth-server-url with the prod profile for Dev Services for Keycloak to start a container. For example:

If a custom realm file has to be imported into Keycloak before running the tests, then you can configure Dev Services for Keycloak as follows:

Finally, write a test code the same way as it is described in the Wiremock section. The only difference is that @QuarkusTestResource is no longer needed:

Use KeycloakTestResourceLifecycleManager for your tests only if there is a good reason not to use Dev Services for Keycloak. If you need to do the integration testing against Keycloak then you are encouraged to do it with Dev Services for Keycloak.

First, add the following dependency:

This provides io.quarkus.test.keycloak.server.KeycloakTestResourceLifecycleManager - an implementation of io.quarkus.test.common.QuarkusTestResourceLifecycleManager which starts a Keycloak container.

Then, configure the Maven Surefire plugin as follows (and similarly the Maven Failsafe plugin when testing in native image):

Now, set the configuration and write the test code the same way as it is described in the Wiremock section. The only difference is the name of QuarkusTestResource:

KeycloakTestResourceLifecycleManager 注册 alice 和 admin 用户。默认情况下，用户 alice 仅具有 user 角色 - 可以使用 keycloak.token.user-roles 系统属性对其进行自定义。默认情况下，用户 admin 具有 用户 和 admin 角色 - 可以使用 keycloak.token.admin-roles 系统属性对其进行自定义。

By default, KeycloakTestResourceLifecycleManager uses HTTPS to initialize a Keycloak instance that can be disabled by specifying keycloak.use.https=false. The default realm name is quarkus and client id is quarkus-web-app - set keycloak.realm and keycloak.web-app.client system properties to customize the values if needed.

You can use @TestSecurity and @OidcSecurity annotations to test the web-app application endpoint code, which depends on either one of the following injections, or all four:

For more information, see Use TestingSecurity with injected JsonWebToken.

To see details about the token verification errors, you must enable io.quarkus.oidc.runtime.OidcProvider TRACE level logging:

To see details about the OidcProvider client initialization errors, enable io.quarkus.oidc.runtime.OidcRecorder TRACE level logging:

From the quarkus dev console, type j to change the application global log level.