Securing Applications and Services Guide

The most important endpoint to understand is the well-known configuration endpoint. It lists endpoints and other configuration options relevant to the OpenID Connect implementation in Keycloak. The endpoint is:

To obtain the full URL, add the base URL for Keycloak and replace {realm-name} with the name of your realm. For example:

http://localhost:8080/realms/master/.well-known/openid-configuration

Some RP libraries retrieve all required endpoints from this endpoint, but for others you might need to list the endpoints individually.

The Authorization Code flow redirects the user agent to Keycloak. Once the user has successfully authenticated with Keycloak, an Authorization Code is created and the user agent is redirected back to the application. The application then uses the authorization code along with its credentials to obtain an Access Token, Refresh Token and ID Token from Keycloak.

The flow is targeted towards web applications, but is also recommended for native applications, including mobile applications, where it is possible to embed a user agent.

For more details refer to the Authorization Code Flow in the OpenID Connect specification.

The Implicit flow works similarly to the Authorization Code flow, but instead of returning an Authorization Code, the Access Token and ID Token is returned. This approach reduces the need for the extra invocation to exchange the Authorization Code for an Access Token. However, it does not include a Refresh Token. This results in the need to permit Access Tokens with a long expiration; however, that approach is not practical because it is very hard to invalidate these tokens. Alternatively, you can require a new redirect to obtain a new Access Token once the initial Access Token has expired. The Implicit flow is useful if the application only wants to authenticate the user and deals with logout itself.

You can instead use a Hybrid flow where both the Access Token and an Authorization Code are returned.

One thing to note is that both the Implicit flow and Hybrid flow have potential security risks as the Access Token may be leaked through web server logs and browser history. You can somewhat mitigate this problem by using short expiration for Access Tokens.

For more details, see the Implicit Flow in the OpenID Connect specification.

Per current OAuth 2.0 Security Best Current Practice, this flow should not be used. This flow is removed from the future OAuth 2.1 specification.

Resource Owner Password Credentials, referred to as Direct Grant in Keycloak, allows exchanging user credentials for tokens. Per current OAuth 2.0 Security Best Practices, this flow should not be used, preferring alternative methods such as Device Authorization Grant or Authorization code.

The limitations of using this flow include:

Security concerns with this flow include:

For a client to be permitted to use the Resource Owner Password Credentials grant, the client has to have the Direct Access Grants Enabled option enabled.

This flow is not included in OpenID Connect, but is a part of the OAuth 2.0 specification. It is removed from the future OAuth 2.1 specification.

For more details, see the Resource Owner Password Credentials Grant chapter in the OAuth 2.0 specification.

Client Credentials are used when clients (applications and services) want to obtain access on behalf of themselves rather than on behalf of a user. For example, these credentials can be useful for background services that apply changes to the system in general rather than for a specific user.

Keycloak provides support for clients to authenticate either with a secret or with public/private keys.

This flow is not included in OpenID Connect, but is a part of the OAuth 2.0 specification.

For more details, see the Client Credentials Grant chapter in the OAuth 2.0 specification.

Device Authorization Grant is used by clients running on internet-connected devices that have limited input capabilities or lack a suitable browser.

For more details, see the OAuth 2.0 Device Authorization Grant specification.

Client Initiated Backchannel Authentication Grant is used by clients who want to initiate the authentication flow by communicating with the OpenID Provider directly without redirect through the user’s browser like OAuth 2.0’s authorization code grant.

The client requests from Keycloak an auth_req_id that identifies the authentication request made by the client. Keycloak creates the auth_req_id.

After receiving this auth_req_id, this client repeatedly needs to poll Keycloak to obtain an Access Token, Refresh Token, and ID Token from Keycloak in return for the auth_req_id until the user is authenticated.

In case that client uses ping mode, it does not need to repeatedly poll the token endpoint, but it can wait for the notification sent by Keycloak to the specified Client Notification Endpoint. The Client Notification Endpoint can be configured in the Keycloak Admin Console. The details of the contract for Client Notification Endpoint are described in the CIBA specification.

For more details, see OpenID Connect Client Initiated Backchannel Authentication Flow specification.

Also refer to other places of Keycloak documentation such as Backchannel Authentication Endpoint of this guide and Client Initiated Backchannel Authentication Grant section of Server Administration Guide. For the details about FAPI CIBA compliance, see the FAPI section of this guide.

Each Java adapter supported by Keycloak can be configured by a simple JSON file. This is what one might look like:

You can use ${…​} enclosure for system property replacement. For example ${jboss.server.config.dir} would be replaced by /path/to/Keycloak. Replacement of environment variables is also supported via the env prefix, for example ${env.MY_ENVIRONMENT_VARIABLE}. You can also define a default value ${…​:default_value}, which is used if the system property ${jboss.server.config.dir:default_value} or the environment variable ${env.MY_ENVIRONMENT_VARIABLE:default_value} is absent.

The initial config file can be obtained from the admin console. This can be done by opening the admin console, select Clients from the menu and clicking on the corresponding client. Once the page for the client is opened click on the Installation tab and select Keycloak OIDC JSON.

Here is a description of each configuration option:

Keycloak provided this adapter in the past, but it is not provided anymore.

We recommend that you switch to the Elytron OIDC library to secure your applications. This library has a similar configuration to the Keycloak WildFly adapters, so you can expect a smooth migration of your applications if you used this adapter with the older Keycloak versions.

Elytron OIDC library works with WildFly 28 or newer versions. For the older WildFly versions or for JBoss EAP 7, it is recommended to upgrade to newer WildFly/EAP or look for some alternative OIDC client library. Otherwise, you will need to stick with the older Keycloak adapters, but those are not maintained and officially supported.

For more details on how to integrate Keycloak with JakartaEE applications running on latest Wildfly/EAP, take a look at the Jakarta EE quickstarts within the Keycloak Quickstart GitHub Repository.

To be able to secure WAR apps deployed on Tomcat 8, and 9, you install the Keycloak Tomcat adapter into your Tomcat installation. You then perform extra configuration to secure each WAR you deploy to Tomcat.

Keycloak has a separate adapter for Jetty 9.4 that you will have to install into your Jetty installation. You then have to provide some extra configuration in each WAR you deploy to Jetty.

To secure an application with Spring Security and Keycloak, add this adapter as a dependency to your project. You then have to provide some extra beans in your Spring Security configuration file and add the Keycloak security filter to your pipeline.

Unlike the other Keycloak Adapters, you should not configure your security in web.xml. However, keycloak.json is still required. In order for Single Sign Out to work properly you have to define a session listener.

If you are deploying your Java Servlet application on a platform where there is no Keycloak adapter you opt to use the servlet filter adapter. This adapter works a bit differently than the other adapters. You do not define security constraints in web.xml. Instead you define a filter mapping using the Keycloak servlet filter adapter to secure the url patterns you want to secure.

In the snippet above there are two url-patterns. /protected/* are the files we want protected, while the /keycloak/* url-pattern handles callbacks from the Keycloak server.

If you need to exclude some paths beneath the configured url-patterns you can use the Filter init-param keycloak.config.skipPattern to configure a regular expression that describes a path-pattern for which the keycloak filter should immediately delegate to the filter-chain. By default no skipPattern is configured.

Patterns are matched against the requestURI without the context-path. Given the context-path /myapp a request for /myapp/index.html will be matched with /index.html against the skip pattern.

Note that you should configure your client in the Keycloak Admin Console with an Admin URL that points to a secured section covered by the filter’s url-pattern.

The Admin URL will make callbacks to the Admin URL to do things like backchannel logout. So, the Admin URL in this example should be http[s]://hostname/{context-root}/keycloak.

If you need to customize the session ID mapper, you can configure the fully qualified name of the class in the Filter init-param keycloak.config.idMapper. Session ID mapper is a mapper that is used to map user IDs and session IDs. By default org.keycloak.adapters.spi.InMemorySessionIdMapper is configured.

The Keycloak filter has the same configuration parameters as the other adapters except you must define them as filter init params instead of context params.

To use this filter, include this maven artifact in your WAR poms:

It’s generally not needed to use JAAS for most of the applications, especially if they are HTTP based, and you should most likely choose one of our other adapters. However, some applications and systems may still rely on pure legacy JAAS solution. Keycloak provides two login modules to help in these situations.

The provided login modules are:

Both modules use the following configuration properties:

Keycloak supports securing desktop (for example Swing, JavaFX) or CLI applications via the KeycloakInstalled adapter by performing the authentication step via the system browser.

The KeycloakInstalled adapter supports a desktop and a manual variant. The desktop variant uses the system browser to gather the user credentials. The manual variant reads the user credentials from STDIN.

The KeycloakSecurityContext interface is available if you need to access to the tokens directly. This could be useful if you want to retrieve additional details from the token (such as user profile information) or you want to invoke a RESTful service that is protected by Keycloak.

In servlet environments it is available in secured invocations as an attribute in HttpServletRequest:

Or, it is available in insecured requests in the HttpSession:

Keycloak has some error handling facilities for servlet based client adapters. When an error is encountered in authentication, Keycloak will call HttpServletResponse.sendError(). You can set up an error-page within your web.xml file to handle the error however you want. Keycloak can throw 400, 401, 403, and 500 errors.

Keycloak also sets a HttpServletRequest attribute that you can retrieve. The attribute name is org.keycloak.adapters.spi.AuthenticationError, which should be cast to org.keycloak.adapters.OIDCAuthenticationError.

For example:

You can log out of a web application in multiple ways. For Jakarta EE servlet containers, you can call HttpServletRequest.logout(). For other browser applications, you can redirect the browser to http://auth-server/realms/{realm-name}/protocol/openid-connect/logout, which logs the user out if that user has an SSO session with his browser. The actual logout is done once the user confirms the logout. You can optionally include parameters such as id_token_hint, post_logout_redirect_uri, client_id and others as described in the OpenID Connect RP-Initiated Logout. As a result, that logout does not need to be explicitly confirmed by the user if you include the id_token_hint parameter. After logout, the user will be automatically redirected to the specified post_logout_redirect_uri as long as it is provided. Note that you need to include either the client_id or id_token_hint parameter in case that post_logout_redirect_uri is included.

If you want to avoid logging out of an external identity provider as part of the logout process, you can supply the parameter initiating_idp, with the value being the identity (alias) of the identity provider in question. This parameter is useful when the logout endpoint is invoked as part of single logout initiated by the external identity provider. The parameter initiating_idp is the supported parameter of the Keycloak logout endpoint in addition to the parameters described in the RP-Initiated Logout specification.

When using the HttpServletRequest.logout() option the adapter executes a back-channel POST call against the Keycloak server passing the refresh token. If the method is executed from an unprotected page (a page that does not check for a valid token) the refresh token can be unavailable and, in that case, the adapter skips the call. For this reason, using a protected page to execute HttpServletRequest.logout() is recommended so that current tokens are always taken into account and an interaction with the Keycloak server is performed if needed.

The Keycloak initial authorization endpoint request has support for various parameters. Most of the parameters are described in OIDC specification. Some parameters are added automatically by the adapter based on the adapter configuration. However, there are also a few parameters that can be added on a per-invocation basis. When you open the secured application URI, the particular parameter will be forwarded to the Keycloak authorization endpoint.

For example, if you request an offline token, then you can open the secured application URI with the scope parameter like:

and the parameter scope=offline_access will be automatically forwarded to the Keycloak authorization endpoint.

The supported parameters are:

Most of the parameters are described in the OIDC specification. The only exception is parameter kc_idp_hint, which is specific to Keycloak and contains the name of the identity provider to automatically use. For more information see the Identity Brokering section in the Server Administration Guide.

When a confidential OIDC client needs to send a backchannel request (for example, to exchange code for the token, or to refresh the token) it needs to authenticate against the Keycloak server. By default, there are three ways to authenticate the client: client ID and client secret, client authentication with signed JWT, or client authentication with signed JWT using client secret.

Multi Tenancy, in our context, means that a single target application (WAR) can be secured with multiple Keycloak realms. The realms can be located on the same Keycloak instance or on different instances.

In practice, this means that the application needs to have multiple keycloak.json adapter configuration files.

You could have multiple instances of your WAR with different adapter configuration files deployed to different context-paths. However, this may be inconvenient and you may also want to select the realm based on something else than context-path.

Keycloak makes it possible to have a custom config resolver so you can choose what adapter config is used for each request.

To achieve this first you need to create an implementation of org.keycloak.adapters.KeycloakConfigResolver. For example:

You also need to configure which KeycloakConfigResolver implementation to use with the keycloak.config.resolver context-param in your web.xml:

This chapter is related to supporting clustered applications deployed to JBoss EAP, WildFly and JBoss AS.

There are a few options available depending on whether your application is:

Dealing with clustering is not quite as simple as for a regular application. Mainly due to the fact that both the browser and the server-side application sends requests to Keycloak, so it’s not as simple as enabling sticky sessions on your load balancer.

The adapter is distributed in several ways, but we recommend that you install the keycloak-js package from NPM:

Alternatively, the library can be retrieved directly from the Keycloak server at /js/keycloak.js and is also distributed as a ZIP archive. We are however considering the inclusion of the adapter directly from the Keycloak server as deprecated, and this functionality might be removed in the future.

One important thing to consider about using client-side applications is that the client has to be a public client as there is no secure way to store client credentials in a client-side application. This consideration makes it very important to make sure the redirect URIs you have configured for the client are correct and as specific as possible.

To use the adapter, create a client for your application in the Keycloak Admin Console. Make the client public by toggling Client authentication to Off on the Capability config page.

You also need to configure Valid Redirect URIs and Web Origins. Be as specific as possible as failing to do so may result in a security vulnerability.

The following example shows how to initialize the adapter. Make sure that you replace the options passed to the Keycloak constructor with those of the client you have configured.

To authenticate, you call the login function. Two options exist to make the adapter automatically authenticate. You can pass login-required or check-sso to the init() function.

You can configure a silent check-sso option. With this feature enabled, your browser will not perform a full redirect to the Keycloak server and back to your application, but this action will be performed in a hidden iframe. Therefore, your application resources are only loaded and parsed once by the browser, namely when the application is initialized and not again after the redirect back from Keycloak to your application. This approach is particularly useful in case of SPAs (Single Page Applications).

To enable the silent check-sso, you provide a silentCheckSsoRedirectUri attribute in the init method. Make sure this URI is a valid endpoint in the application; it must be configured as a valid redirect for the client in the Keycloak Admin Console:

The page at the silent check-sso redirect uri is loaded in the iframe after successfully checking your authentication state and retrieving the tokens from the Keycloak server. It has no other task than sending the received tokens to the main application and should only look like this:

Remember that this page must be served by your application at the specified location in silentCheckSsoRedirectUri and is not part of the adapter.

To enable login-required set onLoad to login-required and pass to the init method:

After the user is authenticated the application can make requests to RESTful services secured by Keycloak by including the bearer token in the Authorization header. For example:

One thing to keep in mind is that the access token by default has a short life expiration so you may need to refresh the access token prior to sending the request. You refresh this token by calling the updateToken() method. This method returns a Promise, which makes it easy to invoke the service only if the token was successfully refreshed and displays an error to the user if it was not refreshed. For example:

By default, the adapter creates a hidden iframe that is used to detect if a Single-Sign Out has occurred. This iframe does not require any network traffic. Instead the status is retrieved by looking at a special status cookie. This feature can be disabled by setting checkLoginIframe: false in the options passed to the init() method.

You should not rely on looking at this cookie directly. Its format can change and it’s also associated with the URL of the Keycloak server, not your application.

By default, the adapter uses the Authorization Code flow.

With this flow, the Keycloak server returns an authorization code, not an authentication token, to the application. The JavaScript adapter exchanges the code for an access token and a refresh token after the browser is redirected back to the application.

Keycloak also supports the Implicit flow where an access token is sent immediately after successful authentication with Keycloak. This flow may have better performance than the standard flow because no additional request exists to exchange the code for tokens, but it has implications when the access token expires.

However, sending the access token in the URL fragment can be a security vulnerability. For example the token could be leaked through web server logs and or browser history.

To enable implicit flow, you enable the Implicit Flow Enabled flag for the client in the Keycloak Admin Console. You also pass the parameter flow with the value implicit to init method:

Note that only an access token is provided and no refresh token exists. This situation means that once the access token has expired, the application has to redirect to Keycloak again to obtain a new access token.

Keycloak also supports the Hybrid flow.

This flow requires the client to have both the Standard Flow and Implicit Flow enabled in the Admin Console. The Keycloak server then sends both the code and tokens to your application. The access token can be used immediately while the code can be exchanged for access and refresh tokens. Similar to the implicit flow, the hybrid flow is good for performance because the access token is available immediately. But, the token is still sent in the URL, and the security vulnerability mentioned earlier may still apply.

One advantage in the Hybrid flow is that the refresh token is made available to the application.

For the Hybrid flow, you need to pass the parameter flow with value hybrid to the init method:

Keycloak supports hybrid mobile apps developed with Apache Cordova. The adapter has two modes for this: cordova and cordova-native:

The default is cordova, which the adapter automatically selects if no adapter type has been explicitly configured and window.cordova is present. When logging in, it opens an InApp Browser that lets the user interact with Keycloak and afterwards returns to the app by redirecting to http://localhost. Because of this behavior, you whitelist this URL as a valid redirect-uri in the client configuration section of the Admin Console.

While this mode is easy to set up, it also has some disadvantages:

Use this example app to help you get started: https://github.com/keycloak/keycloak/tree/master/examples/cordova

The alternative mode is`cordova-native`, which takes a different approach. It opens the login page using the system’s browser. After the user has authenticated, the browser redirects back into the application using a special URL. From there, the Keycloak adapter can finish the login by reading the code or token from the URL.

You can activate the native mode by passing the adapter type cordova-native to the init() method:

This adapter requires two additional plugins:

The technical details for linking to an app differ on each platform and special setup is needed. Please refer to the Android and iOS sections of the deeplinks plugin documentation for further instructions.

Different kinds of links exist for opening apps: * custom schemes, such as myapp://login or android-app://com.example.myapp/https/example.com/login * Universal Links (iOS)) / Deep Links (Android). While the former are easier to set up and tend to work more reliably, the latter offer extra security because they are unique and only the owner of a domain can register them. Custom-URLs are deprecated on iOS. For best reliability, we recommend that you use universal links combined with a fallback site that uses a custom-url link.

Furthermore, we recommend the following steps to improve compatibility with the adapter:

There is an example app that shows how to use the native-mode: https://github.com/keycloak/keycloak/tree/master/examples/cordova-native

In some situations, you may need to run the adapter in environments that are not supported by default, such as Capacitor. To use the JavasScript client in these environments, you can pass a custom adapter. For example, a third-party library could provide such an adapter to make it possible to reliably run the adapter:

This specific package does not exist, but it gives a pretty good example of how such an adapter could be passed into the client.

It’s also possible to make your own adapter, to do so you will have to implement the methods described in the KeycloakAdapter interface. For example the following TypeScript code ensures that all the methods are properly implemented:

Naturally you can also do this without TypeScript by omitting the type information, but ensuring implementing the interface properly will then be left entirely up to you.

In the latest versions of some browsers, various cookies policies are applied to prevent tracking of the users by third parties, such as SameSite in Chrome or completely blocked third-party cookies. Those policies are likely to become more restrictive and adopted by other browsers over time. Eventually cookies in third-party contexts may become completely unsupported and blocked by the browsers. As a result, the affected adapter features might ultimately be deprecated.

The adapter relies on third-party cookies for Session Status iframe, silent check-sso and partially also for regular (non-silent) check-sso. Those features have limited functionality or are completely disabled based on how restrictive the browser is regarding cookies. The adapter tries to detect this setting and reacts accordingly.

Assuming you’ve already installed Node.js, create a folder for your application:

Use npm init command to create a package.json for your application. Now add the Keycloak connect adapter in the dependencies list:

In the root directory of your project create a file called server.js and add the following code:

Install the express-session dependency:

To start the server.js script, add the following command in the 'scripts' section of the package.json:

Now we have the ability to run our server with following command:

By default, this will locate a file named keycloak.json alongside the main executable of your application, in our case on the root folder, to initialize Keycloak specific settings such as public key, realm name, various URLs.

In that case a Keycloak deployment is necessary to access Keycloak admin console.

Please visit links on how to deploy a Keycloak admin console with Podman or Docker

Now we are ready to obtain the keycloak.json file by visiting the Keycloak Admin Console → clients (left sidebar) → choose your client → Installation → Format Option → Keycloak OIDC JSON → Download

Paste the downloaded file on the root folder of our project.

Instantiation with this method results in all the reasonable defaults being used. As alternative, it’s also possible to provide a configuration object, rather than the keycloak.json file:

Applications can also redirect users to their preferred identity provider by using:

Once instantiated, install the middleware into your connect-capable app:

In order to do so, first we have to install Express:

then require Express in our project as outlined below:

and configure Keycloak middleware in Express, by adding at the code below:

Last but not least, let’s set up our server to listen for HTTP requests on port 3000 by adding the following code to main.js:

If the application is running behind a proxy that terminates an SSL connection Express must be configured per the express behind proxies guide. Using an incorrect proxy configuration can result in invalid redirect URIs being generated.

Example configuration:

To secure a resource with an application role for a different app:

To secure a resource with a realm role:

The keycloak-enforcer method operates in two modes, depending on the value of the response_mode configuration option.

If response_mode is set to token, permissions are obtained from the server on behalf of the subject represented by the bearer token that was sent to your application. In this case, a new access token is issued by Keycloak with the permissions granted by the server. If the server did not respond with a token with the expected permissions, the request is denied. When using this mode, you should be able to obtain the token from the request as follows:

Prefer this mode when your application is using sessions and you want to cache previous decisions from the server, as well automatically handle refresh tokens. This mode is especially useful for applications acting as a client and resource server.

If response_mode is set to permissions (default mode), the server only returns the list of granted permissions, without issuing a new access token. In addition to not issuing a new token, this method exposes the permissions granted by the server through the request as follows:

Regardless of the response_mode in use, the keycloak.enforcer method will first try to check the permissions within the bearer token that was sent to your application. If the bearer token already carries the expected permissions, there is no need to interact with the server to obtain a decision. This is specially useful when your clients are capable of obtaining access tokens from the server with the expected permissions before accessing a protected resource, so they can use some capabilities provided by Keycloak Authorization Services such as incremental authorization and avoid additional requests to the server when keycloak.enforcer is enforcing access to the resource.

By default, the policy enforcer will use the client_id defined to the application (for instance, via keycloak.json) to reference a client in Keycloak that supports Keycloak Authorization Services. In this case, the client can not be public given that it is actually a resource server.

If your application is acting as both a public client(frontend) and resource server(backend), you can use the following configuration to reference a different client in Keycloak with the policies that you want to enforce:

It is recommended to use distinct clients in Keycloak to represent your frontend and backend.

If the application you are protecting is enabled with Keycloak authorization services and you have defined client credentials in keycloak.json, you can push additional claims to the server and make them available to your policies in order to make decisions. For that, you can define a claims configuration option which expects a function that returns a JSON with the claims you want to push:

For more details about how to configure Keycloak to protected your application resources, please take a look at the Authorization Services Guide.

Advanced Login Configuration:

By default, all unauthorized requests will be redirected to the Keycloak login page unless your client is bearer-only. However, a confidential or public client may host both browsable and API endpoints. To prevent redirects on unauthenticated API requests and instead return an HTTP 401, you can override the redirectToLogin function.

For example, this override checks if the URL contains /api/ and disables login redirects:

When the user-triggered logout is invoked a query parameter redirect_url can be passed:

This parameter is then used as the redirect url of the OIDC logout endpoint and the user will be redirected to https://example.com/logged/out.

A complete example using the Node.js adapter usage can be found in Keycloak quickstarts for Node.js

To make sure that your clients are FAPI compliant, you can configure Client Policies in your realm as described in the Server Administration Guide and link them to the global client profiles for FAPI support, which are automatically available in each realm. You can use either fapi-1-baseline or fapi-1-advanced profile based on which FAPI profile you need your clients to conform with. You can use also profiles fapi-2-security-profile or fapi-2-message-signing for the compliance with FAPI 2 Draft specifications.

In case you want to use Pushed Authorization Request (PAR), it is recommended that your client use both the fapi-1-baseline profile and fapi-1-advanced for PAR requests. Specifically, the fapi-1-baseline profile contains pkce-enforcer executor, which makes sure that client use PKCE with secured S256 algorithm. This is not required for FAPI Advanced clients unless they use PAR requests.

In case you want to use CIBA in a FAPI compliant way, make sure that your clients use both fapi-1-advanced and fapi-ciba client profiles. There is a need to use the fapi-1-advanced profile, or other client profile containing the requested executors, as the fapi-ciba profile contains just CIBA-specific executors. When enforcing the requirements of the FAPI CIBA specification, there is a need for more requirements, such as enforcement of confidential clients or certificate-bound access tokens.

Keycloak is compliant with the Open Finance Brasil Financial-grade API Security Profile 1.0 Implementers Draft 3. This one is stricter in some requirements than the FAPI 1 Advanced specification and hence it may be needed to configure Client Policies in the more strict way to enforce some of the requirements. Especially:

As confidential information is being exchanged, all interactions shall be encrypted with TLS (HTTPS). Moreover, there are some requirements in the FAPI specification for the cipher suites and TLS protocol versions used. To match these requirements, you can consider configure allowed ciphers. This configuration can be done by setting the https-protocols and https-cipher-suites options. Keycloak uses TLSv1.3 by default and hence it is possibly not needed to change the default settings. However it may be needed to adjust ciphers if you need to fall back to lower TLS version for some reason. For more details, see Configuring TLS guide.

If you need to manually validate access tokens issued by Keycloak, you can invoke the Introspection Endpoint. The downside to this approach is that you have to make a network invocation to the Keycloak server. This can be slow and possibly overload the server if you have too many validation requests going on at the same time. Keycloak issued access tokens are JSON Web Tokens (JWT) digitally signed and encoded using JSON Web Signature (JWS). Because they are encoded in this way, you can locally validate access tokens using the public key of the issuing realm. You can either hard code the realm’s public key in your validation code, or lookup and cache the public key using the certificate endpoint with the Key ID (KID) embedded within the JWS. Depending on what language you code in, many third party libraries exist and they can help you with JWS validation.

When using the redirect based flows, be sure to use valid redirect uris for your clients. The redirect uris should be as specific as possible. This especially applies to client-side (public clients) applications. Failing to do so could result in:

In production for web applications always use https for all redirect URIs. Do not allow redirects to http.

A few special redirect URIs also exist:

Each SAML client adapter supported by Keycloak can be configured by a simple XML text file. This is what one might look like:

Some of these configuration switches may be adapter specific and some are common across all adapters. For Java adapters you can use ${…​} enclosure as System property replacement. For example ${jboss.server.config.dir}.

To be able to secure WAR apps deployed on JBoss EAP or WildFly, you must install and configure the Keycloak SAML Adapter Subsystem.

You then provide a keycloak config, /WEB-INF/keycloak-saml.xml file in your WAR and change the auth-method to KEYCLOAK-SAML within web.xml.

For the WildFly 29 or newer, the SAML adapter is provided as a Galleon feature pack. More details about this is in the WildFly documentation. The same option will be provided for JBoss EAP 8 GA.

Keycloak provided adapter ZIP download in the past, but it is not provided anymore. For the older WildFly versions, it is recommended to upgrade to newer WildFly/EAP and use Galleon. Otherwise, you will need to stick with the older Keycloak adapters, but those are not maintained and officially supported. For JBoss EAP 7, it is possible to stick with the Red Hat Single Sign-On 7.6 adapters, which is still supported.

For more details on how to integrate Keycloak with JakartaEE applications running on latest Wildfly/EAP, take a look at the Jakarta EE quickstart in the Keycloak Quickstart GitHub Repository.

Below are the details on how to configure the SAML adapter secured by Galleon.

To be able to secure WAR apps deployed on Tomcat 8 or 9 you must install the Keycloak Tomcat SAML adapter into your Tomcat installation. You then have to provide some extra configuration in each WAR you deploy to Tomcat.

To be able to secure WAR apps deployed on Jetty you must install the Keycloak Jetty 9.4 SAML adapter into your Jetty installation. You then provide some extra configuration in each WAR you deploy to Jetty.

Use the following installation and configuration procedures.

If you want to use SAML with a Java servlet application that doesn’t have an adapter for that servlet platform, you can opt to use the servlet filter adapter that Keycloak has. This adapter works a little differently than the other adapters. You still have to specify a /WEB-INF/keycloak-saml.xml file as defined in the General Adapter Config section, but you do not define security constraints in web.xml. Instead you define a filter mapping using the Keycloak servlet filter adapter to secure the url patterns you want to secure.

The Keycloak filter has the same configuration parameters available as the other adapters except you must define them as filter init params instead of context params.

You can define multiple filter mappings if you have various different secure and unsecure url patterns.

When registering SPs with an IdP, you must register http[s]://hostname/{context-root}/saml as your Assert Consumer Service URL and Single Logout Service URL.

To use this filter, include this maven artifact in your WAR poms:

In order to use Multi Tenancy the keycloak.config.resolver parameter should be passed as a filter parameter.

For each servlet-based adapter, the endpoint you register for the assert consumer service URL and single logout service must be the base URL of your servlet application with /saml appended to it, that is, https://example.com/contextPath/saml.

There are multiple ways you can log out from a web application. For Jakarta EE servlet containers, you can call HttpServletRequest.logout(). For any other browser application, you can point the browser at any url of your web application that has a security constraint and pass in a query parameter GLO, i.e. http://myapp?GLO=true. This will log you out if you have an SSO session with your browser.

After a successful SAML login, your application code may want to obtain attribute values passed with the SAML assertion. HttpServletRequest.getUserPrincipal() returns a Principal object that you can typecast into a Keycloak specific class called org.keycloak.adapters.saml.SamlPrincipal. This object allows you to look at the raw assertion and also has convenience functions to look up attribute values.

Keycloak has some error handling facilities for servlet based client adapters. When an error is encountered in authentication, the client adapter will call HttpServletResponse.sendError(). You can set up an error-page within your web.xml file to handle the error however you want. The client adapter can throw 400, 401, 403, and 500 errors.

The client adapter also sets an HttpServletRequest attribute that you can retrieve. The attribute name is org.keycloak.adapters.spi.AuthenticationError. Typecast this object to: org.keycloak.adapters.saml.SamlAuthenticationError. This class can tell you exactly what happened. If this attribute is not set, then the adapter was not responsible for the error code.

The best way to troubleshoot problems is to turn on debugging for SAML in both the client adapter and Keycloak Server. Using your logging framework, set the log level to DEBUG for the org.keycloak.saml package. Turning this on allows you to see the SAML requests and response documents being sent to and from the server.

SAML offers Multi Tenancy, meaning that a single target application (WAR) can be secured with multiple Keycloak realms. The realms can be located on the same Keycloak instance or on different instances.

To do this, the application must have multiple keycloak-saml.xml adapter configuration files.

While you could have multiple instances of your WAR with different adapter configuration files deployed to different context-paths, this may be inconvenient and you may also want to select the realm based on something other than context-path.

Keycloak makes it possible to have a custom config resolver, so you can choose which adapter config is used for each request. In SAML, the configuration is only interesting in the login processing; once the user is logged in, the session is authenticated and it does not matter if the keycloak-saml.xml returned is different. For that reason, returning the same configuration for the same session is the correct way to go.

To achieve this, create an implementation of org.keycloak.adapters.saml.SamlConfigResolver. The following example uses the Host header to locate the proper configuration and load it and the associated elements from the applications' Java classpath:

You must also configure which SamlConfigResolver implementation to use with the keycloak.config.resolver context-param in your web.xml:

There are two hosts involved:

All of the following steps need to performed on $sp_host with root privileges.

Browsers are planning to set the default value for the SameSite attribute for cookies to Lax. This setting means that cookies will be sent to applications only if the request originates in the same domain. This behavior can affect the SAML POST binding which may become non-functional. To preserve full functionality of the mod_auth_mellon module, we recommend setting the SameSite value to None for the cookie created by mod_auth_mellon. Not doing so may result in an inability to login using Keycloak.

To set the SameSite value to None, add the following configuration to <Location / > tag within your mellon.conf file.

The support for this configuration is available in the mod_auth_mellon module from version 0.16.0.