Mapping OIDC tokens to schema

You might find that you want to add an identity source to a policy store and map provider claims, or tokens, to your policy store schema. You can automate this process, by using the Guided setup to create your policy store with an identity source, or update your schema manually after the policy store is created. Once you have mapped the tokens to the schema you can create policies that reference them.

This section of the user guide has the following information:

API-linked policy stores and policy stores with an identity source that were created through Guided setup don't require manual mapping of identity (ID) token attributes to schema. You can provide Verified Permissions with the attributes in your user pool and create a schema that is populated with user attributes. In ID token authorization, Verified Permissions maps claims to attributes of a principal entity.

To use an OIDC identity provider (IdP) as an identity source in your Verified Permissions policy store, you must have provider attributes in your schema. The schema is fixed and must correspond to the entities that provider tokens create in IsAuthorizedWithToken or BatchIsAuthorizedWithToken API requests. If you created your policy store in a way that automatically populates your schema from provider information in an ID token, you're ready to write policies. If you create a policy store without a schema for your identity source, you must add provider attributes to the schema that match the entities created using API requests. Then you can write policies using attributes from the provider token.

Mapping ID tokens to schema

Verified Permissions processes ID token claims as the attributes of the user: their names and titles, their group membership, their contact information. ID tokens are most useful in an attribute-based access control (ABAC) authorization model. When you want Verified Permissions to analyze access to resources based on who's making the request, choose ID tokens for your identity source.

Working with ID tokens from an OIDC provider is much the same as working with Amazon Cognito ID tokens. The difference is in the claims. Your IdP might present standard OIDC attributes, or have a custom schema. When you create a new policy store in the Verified Permissions console, you can add an OIDC identity source with an example ID token, or you can manually map token claims to user attributes. Because Verified Permissions isn't aware of the attribute schema of your IdP, you must provide this information.

For more information, see Creating Verified Permissions policy stores.

The following is an example schema for a policy store with an OIDC identity source.

"User": {
   "shape": {
      "type": "Record",
      "attributes": {
         "email": {
            "type": "String"
         },
         "email_verified": {
            "type": "Boolean"
         },
         "name": {
            "type": "String",
            "required": true
         },
         "phone_number": {
            "type": "String"
         },
         "phone_number_verified": {
            "type": "Boolean"
         }
      }
   }
}

For an example policy that will validate against this schema, see Reflects OIDC ID token attributes.

Mapping access tokens

Verified Permissions processes access-token claims other than the groups claim as attributes of the action, or context attributes. Along with group membership, the access tokens from your IdP might contain information about API access. Access tokens are useful in authorization models that use role-based access control (RBAC). Authorization models that rely on access-token claims other than group membership require additional effort in schema configuration.

Most access tokens from external OIDC providers align closely with Amazon Cognito access tokens. An OIDC access token is mapped to a context object when passed to Verified Permissions. Attributes of the access token can be referenced using context.token.attribute_name. The following example OIDC access token includes example base claims.

{
    "sub": "91eb4550-9091-708c-a7a6-9758ef8b6b1e",
    "groups": [
        "Store-Owner-Role",
        "Customer"
    ],
    "iss": "https://auth.example.com",
    "client_id": "1example23456789",
    "aud": "https://myapplication.example.com"
    "scope": "MyAPI-Read",
    "exp": 1688096566,
    "iat": 1688092966,
    "jti": "a1b2c3d4-e5f6-a1b2-c3d4-TOKEN2222222",
    "username": "alice"
}

The following example shows how to reflect the attributes from the example access token in your Verified Permissions schema. For more information about editing your schema, see Editing policy store schemas.

{
   "MyApplication": {
      "actions": {
         "Read": {
            "appliesTo": {
               "context": {
                  "type": "ReusedContext"
               },
               "resourceTypes": [
                  "Application"
               ],
               "principalTypes": [
                  "User"
               ]
            }
         }
      },
      ...
      ...
      "commonTypes": {
         "ReusedContext": {
            "attributes": {
               "token": {
                  "type": "Record",
                  "attributes": {
                     "scope": {
                        "type": "Set",
                        "element": {
                           "type": "String"
                        }
                     },
                     "client_id": {
                        "type": "String"
                     }
                  }
               }
            },
            "type": "Record"
         }
      }
   }
}

For an example policy that will validate against this schema, see Reflects OIDC access token attributes.

Things to know about schema mapping

Attribute mapping differs between token types

In access token authorization, Verified Permissions maps claims to context. In ID token authorization, Verified Permissions maps claims to principal attributes. For policy stores that you create in the Verified Permissions console, only empty and sample policy stores leave you with no identity source and require you to populate your schema with user pool attributes for ID token authorization. Access token authorization is based on role-based access control (RBAC) with group-membership claims and doesn't automatically map other claims to the policy store schema.

Identity source attributes aren't required

When you create an identity source in the Verified Permissions console, no attributes are marked as required. This prevents missing claims from causing validation errors in authorization requests. You can set attributes to required as needed, but they must be present in all authorization requests.

RBAC doesn't require attributes in schema

Schemas for identity sources depend on the entity associations that you make when you add your identity source. An identity source maps one claim to a user entity type, and one claim to a group entity type. These entity mappings are the core of an identity-source configuration. With this minimum information, you can write policies that perform authorization actions for specific users and specific groups that users might be members of, in a role-based access control (RBAC) model. The addition of token claims to the schema extends the authorization scope of your policy store. User attributes from ID tokens have information about users that can contribute to attribute-based access control (ABAC) authorization. Context attributes from access tokens have information like OAuth 2.0 scopes that can contribute additional access-control information from your provider, but require additional schema modifications.

The Set up with API Gateway and an identity provider and Guided setup options in the Verified Permissions console assign ID token claims to the schema. This isn't the case for access token claims. To add non-group access-token claims to your schema, you must edit your schema in JSON mode and add commonTypes attributes. For more information, see Mapping access tokens.

OIDC groups claim supports multiple formats

When you add an OIDC provider, you can choose the name of the groups claim in ID or access tokens that you want to map to a user’s group membership in your policy store. Verified permissions recognizes groups claims in the following formats:

Choose a token type

The way that your policy store works with your identity source depends on a key decision in identity-source configuration: whether you will process ID or access tokens. With an OIDC provider, you must choose a token type when you add the identity source. You can choose ID or access token, and your choice excludes the unchosen token type from being processed in your policy store. Especially if you wish to benefit from the automatic mapping of ID token claims to attributes in the Verified Permissions console, decide early about the token type that you want to process before you create your identity source. Changing the token type requires significant effort to refactor your policies and schema. The following topics describe the use of ID and access tokens with policy stores.

Cedar parser requires brackets for some characters

Policies typically reference schema attributes in a format like principal.username. In the case of most non-alphanumeric characters like :, ., or / that might appear in token claim names, Verified Permissions can't parse a condition value like principal.cognito:username or context.ip-address. You must instead format these conditions with bracket notation in the format principal["cognito:username"] or context["ip-address"], respectively. The underscore character _ is a valid character in claim names, and the only non-alphanumeric exception to this requirement.

A partial example schema for a principal attribute of this type looks like the following:

"User": {
   "shape": {
      "type": "Record",
      "attributes": {
         "cognito:username": {
            "type": "String",
            "required": true
         },
         "custom:employmentStoreCode": {
            "type": "String",
            "required": true,
         },
         "email": {
            "type": "String",
            "required": false
         }
      }
   }
}

A partial example schema for a context attribute of this type looks like the following:

"GetOrder": {
   "memberOf": [],
   "appliesTo": {
      "resourceTypes": [
         "Order"
      ],
      "context": {
         "type": "Record",
         "attributes": {
            "ip-address": {
               "required": false,
               "type": "String"
            }
		 }
	  },
      "principalTypes": [
         "User"
      ]
   }
}

For an example policy that will validate against this schema, see Uses bracket notation to reference token attributes.