Introduction

The OAuth 2.0 authorization framework enables a third-party application (the client) to obtain limited access to an HTTP service on behalf of a resource owner (the end user). For instance, a tool for generating advanced meta analysis graphs (the client) could obtain access to read and possibly modify a review author's (the end user's) reviews in Archie.

The full specification of the OAuth 2.0 authorization framework can be found at: http://tools.ietf.org/html/rfc6749. In the following this is referred to as "the specification". Any features in this document that do not follow the specification are marked "non-standard".

Archie is implementing three of the four authorization methods or 'flows' described in the specification:

Authorization Code Grant flow (section 4.1 of the specification), informally known as the server-side flow.
Implicit Grant flow (section 4.2 of the specification), informally known as the client-side flow.
Client Credentials Grant (section 4.4 of the specification), also known as the client credentials flow.

The Resource Owner Password Credentials Grant is not currently supported. The details and diagrams of each flow are available in the specification.

The aim of each supported flow is for the client to obtain an access token that can be used in subsequent API calls to read and/or modify resources in Archie or Review DB in the context of the end user. This is done in a way (using HTTP redirects) where the end user never has to enter his or her Cochrane Account password into the client application but only on the common Archie login page.

OAuth 2.0 can also be used for single sign on, i.e. the ability for the end user to log into the client using his or her Cochrane Account, by utilising the authentication part of the OAuth flow only, and not using the access token for further API calls.

Registration

Before a client can use the system it must be registered with Archie with the following information:

Client ID;
Client name (for display);
Client secret (password - for confidential clients only, see below);
One or more redirect URIs (to which the end user's user agent (typically a browser) is redirected after completing the Cochrane Account authentication);
A URI pointing to an image/logo that can be displayed on the login/consent page (note: currently not used on the Cochrane Account login page);
A URI pointing to a landing page for the client. This is used on various Cochrane Account pages (except login) to return the user to the client;
The resource scopes that the client should have access to (see below);
A client type.

In addition, for the client credentials flow, an Archie user must be associated with the client. The access token obtained with this flow will be associated with this user and the user's permissions.

There are two different types of client, confidential and public:

A confidential client is one that is capable of keeping its client credentials secret and the access token hidden from the end user. This typically means that a server has to be involved where the access token can be stored in a user session object. A confidential client is using the server-side flow to obtain the access token server-to-server without involving the browser, and in this process it authenticates itself to Archie using its client ID and client secret. Since this is the most secure type of client it has more privileges than a public client, e.g. the ability to refresh access tokens that are about to expire. Confidential clients may also use the client credentials flow to obtain an access token server-to-server without involving an end user.
The typical example of a public client is a pure HTML5/JavaScript app. Since the source code and memory can be inspected by the end user in the browser this type of client is not capable of keeping the access token hidden from the end user. And since the client secret would be visible in the source code there's no point in assigning a secret to a public client. A public client is using the client-side flow to obtain short lived (e.g. 4 hours) access tokens.

The registration of redirect URIs is vital for the security of the system, and Archie only allows complete URIs (no wildcards) to be registered. However, a client is allowed to register multiple URIs.

Endpoints

All endpoints require a secure connection (HTTPS). If plain HTTP is used an error status code 403 (forbidden) is returned.

For the test version of each endpoint, replace archie.cochrane.org with test-archie.cochrane.org.

Primary endpoints

Authorization endpoint:

https://archie.cochrane.org/oauth2/auth

Token endpoint:

https://archie.cochrane.org/oauth2/token

Secondary endpoints (non-standard)

Token information endpoint:

https://archie.cochrane.org/oauth2/tokeninfo

Token revoking endpoint:

https://archie.cochrane.org/oauth2/revoke

Authorization Code Grant (server-side) flow

Step A: Authorization code grant

The server-side flow is optimised for confidential clients, although in theory it could be use by a public client. The first step is for the client to obtain an authorization code from Archie.

The client directs the end user's browser (e.g. in a pop-up window) to the authorization endpoint with the following query parameters ("application/x-www-form-urlencoded" format) added to the endpoint URI:

response_type (required): Must be set to "code" for this flow.

client_id (required): The ID obtained when registering the client.

redirect_uri (optional): A redirect URI registered for the client. Required if multiple URIs have been registered, otherwise defaults to the single registered URI.

state (highly recommended): An (often randomly) generated value used by the client to maintain state between the request and callback. The parameter will be returned unchanged in the callback.

scope (optional): The scope of the resources that the access token should provide access to. Possible values are:

"none": The default, single sign on only, no access to data
"person: Manage person records
"group": Manage groups
"document": Manage documents and reviews
"workflow": Manage tasks and workflows
"crs": Manage studies in CRS
"crso": Manage studies in CRSO
"linked_data": Access to linked data resources
"all": Manage any resource type

Multiple scopes can be provided as a space separated list, e.g. "person document", but "all" or "none" must stand alone. If no scope is specified, "none" is used (note: the default is not all scopes registered with the client). Only scopes registered with the client are allowed.

access_type (optional, non-standard): Possible values are "online" (default) and "offline". If set to "offline" for a confidential client, a refresh token is issued together with the access token in step B.

At the authorization endpoint the end user will have to log into Cochrane Account if he or she is not logged in already. After the authentication, provided that the scope of the request is different from "none", the end user may be presented with a consent screen where he or she has to agree to give the client access to resources in Archie within the given scope before the flow may continue. Note: the consent screen is not shown for clients hosted on *.cochrane.org domains.

Once the authentication and possible consent is in order, i.e. the end user has authorized the client to access his or her data, the browser is redirected back to the redirect URI (using a HTTP 302 status code) with the following query parameters added:

code: The authorization code generated by Archie.

state (optional): The state parameter provided by the client, if any.

In case of an error condition, depending on the error type, an error message will be displayed either directly to the end user (if there is a problem with the client ID or redirect URI), or it will be provided as an error parameter in the redirect URI to the client. For instance, if the end user does not give his or her consent, an "access_denied" error is returned to the client. See the specification section 4.1.2.1 for details.

Step B: Access token request

In the second step the client exchanges the authorization code with an access token server-to-server without involving the user's browser.

The client's server makes a POST request to the token endpoint with the following parameters included ("application/x-www-form-urlencoded" format):

grant_type (required): Must be set to "authorization_code" for this flow.

code (required): The authorization code obtained in step A.

redirect_uri (optional): Required if the redirect_uri parameter was provided in step A, in which case the parameter must have the exact same value. This parameter is not used for redirecting at this step, only for added security.

Client authentication

For any call to the token endpoint, including this, the client must provide its credentials as Basic HTTP authentication with client ID as user name and client secret as password. The authentication must be provided in the Authorization header for a confidential client.

Note: the specification allows, but discourages, providing the information as parameters in the request body instead (client_id and client_secret), but Archie does not allow this for confidential clients. And since a public client is not expected to use this endpoint and this flow, this is not relevant for practical purposes.

Before building the header for Basic authentication, client IDs and client secrets containing special characters including %, +, and all non-ASCII characters must be encoded using the UTF-8 character encoding scheme first; the resulting octet sequence then needs to be further application/x-www-form-urlencoded (see section 2.3.1 of the specification). To build the header, the encoded credentials are separated by a colon, the resulting string is base 64 encoded, and the result is prefixed with "Basic ".

Access code response

The response to the POST request is a JSON object (MIME type "application/json") like this:

{
    "access_token": "3e8ec1a3d43c983b57df0616b498c04807b466e919999aa0f3f3aabca1dd48cc",
    "token_type": "Bearer",
    "expires_in": "14400",
    "refresh_token": "fa5e931aa4cecfcf4ae3bebd9648aa14942d6c79f6c26ffa3bd44d1e106522bc"
}

The first field is the actual access token which should be kept on the server and never passed to the end user's browser. The token_type will always be "Bearer" in Archie's implementation. The number of seconds that the token is valid is given in expires_in. A refresh_token field is only returned if the access_type in step A is set to "offline".

Error response

In case of an error condition Archie responds with a HTTP 400 status code with the error message included in the body of the response as a JSON object, e.g.:

{  
    "error": "invalid_grant",
    "error_description": "Invalid authorization code."
}

For further details see section 5.2 of the specification.

Implicit Grant (client-side) flow

The client-side flow is optimised for public clients such as pure HTML5/JavaScript clients. In fact, Archie does not permit a confidential client to use this flow (non-standard?).

The client directs the end user's browser (e.g. in a pop-up window or IFRAME) to the authorization endpoint with the following query parameters added ("application/x-www-form-urlencoded" format) to the endpoint URI:

response_type (required): Must be set to "token" for this flow.

client_id (required): The ID obtained when registering the client.

redirect_uri (optional): A redirect URI registered for the client. Required if multiple URIs have been registered, otherwise defaults to the single registered URI.

state (highly recommended): A (often randomly) generated value used by the client to maintain state between the request and callback. The parameter will be returned unchanged in the callback.

scope (optional): The scope of the resources that the access token should provide access to. Possible values are:

"none": The default, single sign on only, no access to data
"person: Manage person records
"group": Manage groups
"document": Manage documents and reviews
"workflow": Manage tasks and workflows
"crs": Manage studies in CRS
"crso": Manage studies in CRSO
"all": Manage any resource type

Multiple scopes can be provided as a space separated list, e.g. "person document", but "all" or "none" must stand alone. If no scope is specified, "none" is used (note: the default is not all scopes registered with the client). Only scopes registered with the client are allowed.

The authentication and consent screen is displayed as for the server-side flow. Once completed, the browser is redirected back to the redirect URI (using a HTTP 302 status code) with the following parameters added to the fragment component of the redirect URI, i.e. after the hash sign:

access_token: The access token issued by Archie.

token_type: Always "Bearer" in Archie's implementation.

expires_in: The number of seconds that the token is valid.

state (optional): The state parameter provided by the client, if any.

In case of an error condition, depending on the error type, an error message will either be displayed directly to the end user (if there is a problem with the client ID or redirect URI), or it will be provided as an error parameter in the fragment component of the redirect URI to the client. See the specification section 4.2.2.1 for details.

Client Credentials Grant flow

The client credentials flow allows a confidential client (server) to obtain an access token server-to-server without involving an end user. Before this flow is used, an Archie user must be associated with the client. The access token obtained with this flow will be associated with the user and the user's permissions.

The server makes a POST request to the token endpoint with the following parameters included ("application/x-www-form-urlencoded" format):

grant_type (required): Must be set to "client_credentials" for this flow.

scope (optional): The scope of the resources that the access token should provide access to. Possible values are:

"none": The default, single sign on only, no access to data
"person: Manage person records
"group": Manage groups
"document": Manage documents and reviews
"workflow": Manage tasks and workflows
"crs": Manage studies in CRS
"crso": Manage studies in CRSO
"linked_data": Access to linked data resources
"all": Manage any resource type

Multiple scopes can be provided as a space separated list, e.g. "person document", but "all" or "none" must stand alone. If no scope is specified, "none" is used (note: the default is not all scopes registered with the client). Only scopes registered with the client are allowed.

Authentication must be provided in the Authorization header as described for the Authorization Code Grant under client authentication.

Access code response

The response to the POST request is a JSON object (MIME type "application/json") like this:

{
    "access_token": "3e8ec1a3d43c983b57df0616b498c04807b466e919999aa0f3f3aabca1dd48cc",
    "token_type": "Bearer",
    "expires_in": "14400"
}

The access_token field contains the actual access token. The token_type will always be "Bearer" in Archie's implementation. The number of seconds that the token is valid is given in expires_in.

Error response

In case of an error condition Archie responds with a HTTP 400 status code with the error message included in the body of the response as a JSON object, e.g.:

{  
    "error": "invalid_grant",
    "error_description": "Invalid authorization code."
}

For further details see section 5.2 of the specification.

Validating the access token (non-standard)

Archie provides a mechanism for validating and obtaining information about an access token. If the OAuth 2.0 authorization framework is used for single sign on only by a public client it is important to validate the token to ensure that the request to the redirect URI is genuine.

To validate a token, make a GET or POST request to the token information endpoint (see above) with the following query parameters included:

access_token (required): The token to be validated. This can also be provided in the Authorization header in the form: Bearer [access token]

include_permissions (optional): Value can be "true" or "false" (default). Specifies whether the response should include information about the end user's Archie permissions within the scope of the access token.

detailed (optional): Value can be "true" or "false" (default). Specifies whether the response should include the following information in the response: user_id, name (combined), first_name, last_name and email.

The response is a JSON object (MIME type "application/json") like this:

{
    "user_name": "testuser",
    "scope": "document",
    "expires_in": "2410",
    "client_id": "my_account",
    "person_id": "11143"
}

Use person_id as a user's stable ID

Note that the user_name for a given user may change over time, e.g. if the user's email changes, and if the user account is closed it may be reused later by another account. If you need an immutable identifier that will not be reused, person_id is recommended.

Or, with both detailed=true and include_permissions=true:

{
    "user_name": "testuser",
    "scope": "person",
    "expires_in": "3192",
    "client_id": "my_account",
    "person_id": "11143",
    "user_id": "08573771072139469457090803092242",
    "name": "Test User",
    "first_name": "Test",
    "last_name": "User",
    "email": "test-user@cochrane.org",
    "permissions": [
        {
            "resource": "person",
            "entity": "D37BDF5182E26AA2013C72096D0CE6B1",
            "grants": ["view", "view_private", "write"]

        }
    ]
}

In case of an error condition the response will follow the same structure as for the token endpoint (see above).

Refreshing an access token

Refresh tokens can be used to obtain a new access token without asking for the end user's consent. Access tokens expire after 4 hours unless they are revoked. Refresh tokens are valid for 180 days or until they are used to obtain a new access token.

To refresh an access token (server-side flow only) the client's server makes a POST request to the token endpoint with the following parameters included ("application/x-www-form-urlencoded" format):

grant_type (required): Must be set to "refresh_token".

refresh_token (required): The previously obtained refresh token.

The format of the response is identical to the one returned when the original access token was obtained:

{
    "access_token": "71c9e4fba2f43815719b82132f3104ee71930031f50087049e3beb357d9a5107",
    "token_type": "Bearer",
    "expires_in": "14400",
    "refresh_token": "abad3cdea83b2e4d28594db35b526689da1e295fb34dcb13668723f1ecac14e7"
}

Note that a new refresh token is also issued with the response, so the old refresh token must be discarded by the client and replaced by the new token.

In case of an error condition, e.g. if the refresh token is no longer valid, the response will follow the same structure as for other token endpoint errors. The client should discard the refresh token and obtain a new access token using the server-side flow.

Revoking an access token (non standard)

An access token can be revoked/invalidated by making a GET or POST request to the token revoking endpoint (see above) with a single parameter:

access_token (required): The token to be revoked.

In case of an error condition the response will follow the same structure as for the token endpoint (see above).

For security reasons public clients are adviced to revoke any tokens that are no longer being used, e.g. when the user logs out of the application.

Using an access token to access the API

The access token must be passed to the API in the Authorization header, e.g.:

GET https://archie.cochrane.org/rest/reviews/CD000004 HTTP/1.1

Authorization: Bearer 3e8ec1a3d43c983b57df0616b498c04807b466e919999aa0f3f3aabca1dd48cc

The result, if the call succeeds, is the review in XML format (MIME type "application/xml"). If the call fails an HTTP error status code will be returned.

See Review Document API for more examples.

Checking permissions for a single review

As another example, the API contains a method to get the user's permissions for a single review:

GET https://archie.cochrane.org/rest/reviews/CD000004/permissions HTTP/1.1

Authorization: Bearer 3e8ec1a3d43c983b57df0616b498c04807b466e919999aa0f3f3aabca1dd48cc

The result, if the call succeeds, is a JSON object (MIME type "application/json") listing the user's permissions, e.g.:

{
 "permissions": ["view", "read_published", "read_to_be_published", "read_author_drafts", "read_shared_drafts", "write_authoring", "view_author_roles"],
 "read": true
 "write": true
 "delete": false
}

The "read", "write" and "delete" properties in the response that take the current phase of the review into account. For instance, "write": true could be returned if the user can write in authoring phase ("write_authoring" permission) AND the review is in authoring phase.

If the call fails an HTTP error status code will be returned.

Page tree

OAuth 2.0 authentication in Keycloack

Introduction

Registration

Endpoints

Primary endpoints

Secondary endpoints (non-standard)

Authorization Code Grant (server-side) flow

Step A: Authorization code grant

Step B: Access token request

Client authentication

Access code response

Error response

Implicit Grant (client-side) flow

Client Credentials Grant flow

Access code response

Error response

Validating the access token (non-standard)

Refreshing an access token

Revoking an access token (non standard)

Using an access token to access the API

Checking permissions for a single review