Authorization Services Guide

Three main processes define the necessary steps to understand how to use Keycloak to enable fine-grained authorization to your applications:

Authorization services consist of the following RESTFul endpoints:

Each of these services provides a specific API covering the different steps involved in the authorization process.

Per OAuth2 terminology, a resource server is the server hosting the protected resources and capable of accepting and responding to protected resource requests.

Resource servers usually rely on some kind of information to decide whether access to a protected resource should be granted. For RESTful-based resource servers, that information is usually carried in a security token, typically sent as a bearer token along with every request to the server. Web applications that rely on a session to authenticate users usually store that information in the user’s session and retrieve it from there for each request.

In Keycloak, any confidential client application can act as a resource server. This client’s resources and their respective scopes are protected and governed by a set of authorization policies.

A resource is part of the assets of an application and the organization. It can be a set of one or more endpoints, a classic web resource such as an HTML page, and so on. In authorization policy terminology, a resource is the object being protected.

Every resource has a unique identifier that can represent a single resource or a set of resources. For instance, you can manage a Banking Account Resource that represents and defines a set of authorization policies for all banking accounts. But you can also have a different resource named Alice’s Banking Account, which represents a single resource owned by a single customer, which can have its own set of authorization policies.

A resource’s scope is a bounded extent of access that is possible to perform on a resource. In authorization policy terminology, a scope is one of the potentially many verbs that can logically apply to a resource.

It usually indicates what can be done with a given resource. Examples of scopes are view, edit, delete, and so on. However, scope can also be related to specific information provided by a resource. In this case, you can have a project resource and a cost scope, where the cost scope is used to define specific policies and permissions for users to access a project’s cost.

Consider this simple and very common permission:

A permission associates the object being protected with the policies that must be evaluated to determine whether access is granted.

Keycloak provides a rich platform for building a range of permission strategies ranging from simple to very complex, rule-based dynamic permissions. It provides flexibility and helps to:

A policy defines the conditions that must be satisfied to grant access to an object. Unlike permissions, you do not specify the object being protected but rather the conditions that must be satisfied for access to a given object (for example, resource, scope, or both). Policies are strongly related to the different access control mechanisms (ACMs) that you can use to protect your resources. With policies, you can implement strategies for attribute-based access control (ABAC), role-based access control (RBAC), context-based access control, or any combination of these.

Keycloak leverages the concept of policies and how you define them by providing the concept of aggregated policies, where you can build a "policy of policies" and still control the behavior of the evaluation. Instead of writing one large policy with all the conditions that must be satisfied for access to a given resource, the policies implementation in Keycloak Authorization Services follows the divide-and-conquer technique. That is, you can create individual policies, then reuse them with different permissions and build more complex policies by combining individual policies.

Policy providers are implementations of specific policy types. Keycloak provides built-in policies, backed by their corresponding policy providers, and you can create your own policy types to support your specific requirements.

Keycloak provides an SPI (Service Provider Interface) that you can use to plug in your own policy provider implementations.

A permission ticket is a special type of token defined by the User-Managed Access (UMA) specification that provides an opaque structure whose form is determined by the authorization server. This structure represents the resources and/or scopes being requested by a client, the access context, as well as the policies that must be applied to a request for authorization data (requesting party token [RPT]).

In UMA, permission tickets are crucial to support person-to-person sharing and also person-to-organization sharing. Using permission tickets for authorization workflows enables a range of scenarios from simple to complex, where resource owners and resource servers have complete control over their resources based on fine-grained policies that govern the access to these resources.

In the UMA workflow, permission tickets are issued by the authorization server to a resource server, which returns the permission ticket to the client trying to access a protected resource. Once the client receives the ticket, it can make a request for an RPT (a final token holding authorization data) by sending the ticket back to the authorization server.

For more information on permission tickets, see User-Managed Access and the UMA specification.

On the Resource Server Settings page, you can configure the policy enforcement mode, allow remote resource management, and export the authorization configuration settings.

You can change the default configuration by removing the default resource, policy, or permission definitions and creating your own.

The default resource is created with a URI that maps to any resource or path in your application using a /* pattern. Before creating your own resources, permissions and policies, make sure the default configuration doesn’t conflict with your own settings.

The configuration file is exported in JSON format and displayed in a text area, from which you can copy and paste. You can also click Download to download the configuration file and save it.

You can import a configuration file for a resource server.

Resources may have attributes associated with them. These attributes can be used to provide additional information about a resource and to provide additional information to policies when evaluating permissions associated with a resource.

Each attribute is a key and value pair where the value can be a set of one or many strings. Multiple values can be defined for an attribute by separating each value with a comma.

The type field of a resource can be used to group different resources together, so they can be protected using a common set of permissions.

Resources also have an owner. By default, resources are owned by the resource server.

However, resources can also be associated with users, so you can create permissions based on the resource owner. For example, only the resource owner is allowed to delete or update a given resource.

Resource management is also exposed through the Protection API to allow resource servers to remotely manage their resources.

When using the Protection API, resource servers can be implemented to manage resources owned by their users. In this case, you can specify the user identifier to configure a resource as belonging to a specific user.

When creating a role-based policy, you can specify a specific role as Required. When you do that, the policy will grant access only if the user requesting access has been granted all the required roles. Both realm and client roles can be configured as such.

To specify a role as required, select the Required checkbox for the role you want to configure as required.

Required roles can be useful when your policy defines multiple roles but only a subset of them are mandatory. In this case, you can combine realm and client roles to enable an even more fine-grained role-based access control (RBAC) model for your application. For example, you can have policies specific for a client and require a specific client role associated with that client. Or you can enforce that access is granted only in the presence of a specific realm role. You can also combine both approaches within the same policy.

Keycloak allows you to deploy a JAR file in order to deploy scripts to the server. Please, take a look at JavaScript Providers for more details.

Once you have your scripts deployed, you should be able to select the scripts you deployed from the list of available policy providers.

Access is only granted if all conditions are satisfied. Keycloak will perform an AND based on the outcome of each condition.

When creating aggregated policies, you can also define the decision strategy that will be used to determine the final decision based on the outcome from each policy.

By default, when you add a group to this policy, access restrictions will only apply to members of the selected group.

Under some circumstances, it might be necessary to allow access not only to the group itself but to any child group in the hierarchy. For any group added you can mark a checkbox Extend to Children in order to extend access to child groups.

In the example above, the policy is granting access for any user member of IT or any of its children.

When creating a client scope-based policy, you can specify a specific client scope as Required. When you do that, the policy will grant access only if the client requesting access has been granted all the required client scopes.

To specify a client scope as required, select the Required checkbox for the client scope you want to configure as required.

Required client scopes can be useful when your policy defines multiple client scopes but only a subset of them are mandatory.

The evaluation context provides useful information to policies during their evaluation.

From this interface, policies can obtain:

The Identity is built based on the OAuth2 Access Token that was sent along with the authorization request, and this construct has access to all claims extracted from the original token. For example, if you are using a Protocol Mapper to include a custom claim in an OAuth2 Access Token you can also access this claim from a policy and use it to build your conditions.

The EvaluationContext also gives you access to attributes related to both the execution and runtime environments. For now, these are only a few built-in attributes.

Resource permissions can also be used to define policies that are to be applied to all resources with a given type. This form of resource-based permission can be useful when you have resources sharing common access requirements and constraints.

Frequently, resources within an application can be categorized (or typed) based on the data they encapsulate or the functionality they provide. For example, a financial application can manage different banking accounts where each one belongs to a specific customer. Although they are different banking accounts, they share common security requirements and constraints that are globally defined by the banking organization. With typed resource permissions, you can define common policies to apply to all banking accounts, such as:

To create a typed resource permission, click Apply to Resource Type when creating a new resource-based permission. With Apply to Resource Type set to On, you can specify the type that you want to protect as well as the policies that are to be applied to govern access to all resources with the type you have specified.

Clients need to authenticate to the token endpoint in order to obtain an RPT. When using the urn:ietf:params:oauth:grant-type:uma-ticket grant type, clients can use any of these authentication methods:

When obtaining permissions from the server you can push arbitrary claims in order to have these claims available to your policies when evaluating permissions.

If you are obtaining permissions from the server without using a permission ticket (UMA flow), you can send an authorization request to the token endpoint as follows:

The claim_token parameter expects a BASE64 encoded JSON with a format similar to the example below:

The format expects one or more claims where the value for each claim must be an array of strings.

In UMA, the authorization process starts when a client tries to access a UMA protected resource server.

A UMA protected resource server expects a bearer token in the request where the token is an RPT. When a client requests a resource at the resource server without an RPT:

The resource server sends a response back to the client with a permission ticket and an as_uri parameter with the location of a Keycloak server to where the ticket should be sent in order to obtain an RPT.

The permission ticket is a special type of token issued by Keycloak Permission API. They represent the permissions being requested (e.g.: resources and scopes) as well any other information associated with the request. Only resource servers are allowed to create those tokens.

Now that the client has a permission ticket and also the location of a Keycloak server, the client can use the discovery document to obtain the location of the token endpoint and send an authorization request.

If Keycloak assessment process results in issuance of permissions, it issues the RPT with which it has associated the permissions:

The response from the server is just like any other response from the token endpoint when using some other grant type. The RPT can be obtained from the access_token response parameter. In case the client is not authorized to have permissions Keycloak responds with a 403 HTTP status code:

As part of the authorization process, clients need first to obtain a permission ticket from a UMA protected resource server in order to exchange it with an RPT at the Keycloak Token Endpoint.

By default, Keycloak responds with a 403 HTTP status code and a request_denied error in case the client can not be issued with an RPT.

Such response implies that Keycloak could not issue an RPT with the permissions represented by a permission ticket.

In some situations, client applications may want to start an asynchronous authorization flow and let the owner of the resources being requested decide whether or not access should be granted. For that, clients can use the submit_request request parameter along with an authorization request to the token endpoint:

When using the submit_request parameter, Keycloak will persist a permission request for each resource to which access was denied. Once created, resource owners can check their account and manage their permissions requests.

You can think about this functionality as a Request Access button in your application, where users can ask other users for access to their resources.

Users can manage access to their resources using the Keycloak Account Console. To enable this functionality, you must first enable User-Managed Access for your realm.

A protection API token (PAT) is a special OAuth2 access token with a scope defined as uma_protection. When you create a resource server, Keycloak automatically creates a role, uma_protection, for the corresponding client application and associates it with the client’s service account.

Resource servers can obtain a PAT from Keycloak like any other OAuth2 access token. For example, using curl:

The example above is using the client_credentials grant type to obtain a PAT from the server. As a result, the server returns a response similar to the following:

Resource servers can manage their resources remotely using a UMA-compliant endpoint.

This endpoint provides operations outlined as follows (entire path omitted for clarity):

For more information about the contract for each of these operations, see UMA Resource Registration API.

Resource servers using the UMA protocol can use a specific endpoint to manage permission requests. This endpoint provides a UMA-compliant flow for registering permission requests and obtaining a permission ticket.

A permission ticket is a special security token type representing a permission request. Per the UMA specification, a permission ticket is:

A correlation handle that is conveyed from an authorization server to a resource server, from a resource server to a client, and ultimately from a client back to an authorization server, to enable the authorization server to assess the correct policies to apply to a request for authorization data.

In most cases, you do not need to deal with this endpoint directly. Keycloak provides a policy enforcer that enables UMA for your resource server so it can obtain a permission ticket from the authorization server, return this ticket to client application, and enforce authorization decisions based on a final requesting party token (RPT).

The process of obtaining permission tickets from Keycloak is performed by resource servers and not regular client applications, where permission tickets are obtained when a client tries to access a protected resource without the necessary grants to access the resource. The issuance of permission tickets is an important aspects when using UMA as it allows resource servers to:

Client wise, a permission ticket has also important aspects that its worthy to highlight:

These are just some of the benefits brought by UMA where other aspects of UMA are strongly based on permission tickets, specially regarding privacy and user controlled access to their resources.

Keycloak leverages the UMA Protection API to allow resource servers to manage permissions for their users. In addition to the Resource and Permission APIs, Keycloak provides a Policy API from where permissions can be set to resources by resource servers on behalf of their users.

The Policy API is available at:

This API is protected by a bearer token that must represent a consent granted by the user to the resource server to manage permissions on his behalf. The bearer token can be a regular access token obtained from the token endpoint using:

Sometimes you might want to introspect a requesting party token (RPT) to check its validity or obtain the permissions within the token to enforce authorization decisions on the resource server side.

There are two main use cases where token introspection can help you:

The token introspection is essentially a OAuth2 token introspection-compliant endpoint from which you can obtain information about an RPT.

To introspect an RPT using this endpoint, you can send a request to the server as follows:

The introspection endpoint expects two parameters:

As a result, the server response is:

If the RPT is not active, this response is returned instead:

No. Just like a regular access token issued by a Keycloak server, RPTs also use the JSON web token (JWT) specification as the default format.

If you want to validate these tokens without a call to the remote introspection endpoint, you can decode the RPT and query for its validity locally. Once you decode the token, you can also use the permissions within the token to enforce authorization decisions.

This is essentially what the policy enforcers do. Be sure to:

If a resource server is protected by a policy enforcer, it responds to client requests based on the permissions carried along with a bearer token. Typically, when you try to access a resource server with a bearer token that is lacking permissions to access a protected resource, the resource server responds with a 401 status code and a WWW-Authenticate header.

See UMA Authorization Process for more information.

What your client needs to do is extract the permission ticket from the WWW-Authenticate header returned by the resource server and use the library to send an authorization request as follows:

The authorize function is completely asynchronous and supports a few callback functions to receive notifications from the server:

Most applications should use the onGrant callback to retry a request after a 401 response. Subsequent requests should include the RPT as a bearer token for retries.

The keycloak-js/authz library provides an entitlement function that you can use to obtain an RPT from the server by providing the resources and scopes your client wants to access.

When using the entitlement function, you must provide the client_id of the resource server you want to access.

The entitlement function is completely asynchronous and supports a few callback functions to receive notifications from the server:

Both authorize and entitlement functions accept an authorization request object. This object can be set with the following properties:

If you have already obtained an RPT using any of the authorization functions provided by the library, you can always obtain the RPT as follows from the authorization object (assuming that it has been initialized by one of the techniques shown earlier):