Current File(s): conf/authn/spnego-authn-config.xml, views/spnego-unavailable.vm, views/user-prefs.vm
Format: Native Spring
The authn/SPNEGO login flow supports SPNEGO-based Kerberos authentication, complying with RFC 4559, "SPNEGO-based Kerberos and NTLM HTTP Authentication" (http://tools.ietf.org/html/rfc4559). (It only supports Kerberos, but not the NTLM protocol.)
This mechanism allows the IdP to authenticate users by verifying a Kerberos service ticket sent by the client. Most current web browsers, including Internet Explorer, Firefox, Chrome, Opera, Safari and Konqueror, support SPNEGO/Kerberos based authentication. SPNEGO/Kerberos is most-often used in Microsoft Windows environments, and typically assumes the client machine is joined to a domain so that Kerberos credentials are obtained automatically. It can be tested, and given more technically-skilled users, used, without a domain-joined machine. It also works with MIT or Heimdal Kerberos, not just AD.
This login flow differs from the password-based Kerberos authentication provided by the authn/Password login flow. Where the authn/Password login flow relies on the password submitted to the IdP, the authn/SPNEGO login flow consumes a Kerberos ticket provided by the client, and the IdP never sees the password.
By default, this flow is configured without support for advanced authentication controls like passive or forced authentication, since this is generally not possible with SPNEGO authentication.
The SPNEGO login flow can be used via "opt-in" mode or "enforced" mode. In "opt-in" mode, users need to enable login via SPNEGO using an auto-login checkbox or button (see below). In "enforced" mode, SPNEGO is always tried (though possibly skipped in some cases based on an activation condition), independent of the auto-login option set by the users. By default, "opt-in" mode is used. The "enforced" mode is recommended only if you can ensure that Kerberos works in most situations for which any attached activation condition applies.
To use the authn/SPNEGO login flow, you need to install and configure Kerberos on your IdP server first. This includes the creation of a service principal in the Kerberos realm for your service, and usually includes obtaining a keytab file for that principal. A service password may also be used. See Kerberos Infrastructure for more information.
General configuration of Kerberos is outside the scope of the IdP, and not described in detail here, but no native Kerberos libraries beyond Oracle's Java implementation are required or used.
To use the authn/SPNEGO login flow, it is necessary to have the Kerberos environment configured and working properly.
Some interesting tutorials that may help are:
Before you start, please check that:
- The service-principal (usually "HTTP/principal@your_realm.com") was configured at the KDC.
- The keytab file holding the key of the service-principal was generated (using a keytab is recommended).
- On the IdP server: Check if it is possible to get the service tickets from the KDC with the command kinit.
The clients' browsers need to be configured to support SPNEGO. See Single sign-on Browser configuration for details.
Use conf/authn/spnego-authn-config.xml to configure this flow.
To enable "opt-in" mode, set the value of the bean shibboleth.authn.SPNEGO.EnforceRun to "FALSE". This is the default. To enable "enforced" mode, set the value to "TRUE".
You also need to configure the Kerberos service principal(s) you want to use in the shibboleth.authn.SPNEGO.Krb5.Realms bean. A usual configuration involves a single realm and service principal. If your environment contains multiple realms, you may need to configure more than one service principal. They will be tried in sequence when attempting to accept a ticket from the client.
Each value of the realms list bean must be a bean inherited from shibboleth.KerberosRealmSettings and identifies the service principal and keytab file or password to use. A keytab is recommended, but is a bit more work to obtain securely from your Kerberos administrator. Service principals used with SPNEGO MUST be of the form "HTTP/hostname", where "hostname" is the FQDN of the IdP service. This is required because the client will request a service ticket for that principal.
The shibboleth.authn.SPNEGO.ClassifiedMessageMap bean is a map of error messages to classified error conditions. The default mapping supplied allows a couple of errors raised by the external SPNEGO implementation to be passed through so that meaningful error messages specific to SPNEGO can be shown (e.g., on the login form of the Password login flow in a fall-through scenario). Usually, you won't need to change this mapping, but if you want other errors to cause the IdP to "fall through" to other login flows, you would need to map them to the "ReselectFlow" key.
Finally, note that a common deployment model is to enable the SPNEGO and the Password login flows together, requiring that both be enabled, usually via the conf/idp.properties file:
Kerberos System Configuration
The actual Kerberos configuration is managed in a krb5.conf or krb5.ini file that can be placed in a number of different locations. On non-Windows servers, using a system-wide configuration in /etc/krb5.conf is generally advised. It's possible to have Java-specific configurations and/or provide the path to a configuration using the system property
java.security.krb5.conf, as discussed here. If you want to set the system property
java.security.krb5.conf, you should set it in the configuration of your Servlet container.
Configuration of an Activation Condition
There are known cases where the SPNEGO login process might break and stop in the browser, without returning to the IdP. As the IdP can't recover from this situation, the user can't continue the login process.
The following case is known:
- Internet Explorer: If Internet Explorer gets a SPNEGO login request, but Kerberos is not available, Internet Explorer will fallback to the (deprecated) NTLM mechanism and show a login box to the user. No response will be sent to the IdP (unless the user actually enters credentials), and the user might get confused. (This case actually covers all Internet Explorer users who are not logged into a Windows Domain.)
The IdP provides a flexible mechanism to help avoid some of these situations, the use of an activation condition attached to the login flow's descriptor bean.
The SPNEGO protocol itself doesn't provide any way for the IdP to reliably decide whether the browser supports a Kerberos login or not. You need to specify the conditions under which a Kerberos login will be available in your environment, typically based on the client's IP address or some text appearing in the user agent's identifier string. It's possible to implement such a condition in Java, but using a script is usually simpler.
An activation condition is configured in the file conf/authn/general-authn.xml as follows (this isn't the only way, but it serves as an example):
- Add a new bean anywhere in the file (but outside the
<util:list id="shibboleth.AvailableAuthenticationFlows">bean) with the ID shibboleth.SPNEGO.ActivationCondition with parent bean shibboleth.Conditions.Scripted.
- By setting the
customObjectproperty of this bean to the bean named shibboleth.HttpServletRequest, the current HTTP request object is available in the variable
custom. This allows accessing various properties of the user agent, e.g., the current IP address of the client or the user agent's identifier string. Examples are included below.
]]>lines). The last expression in your code should evaluate to
Add the activation condition to the flow descriptor bean with ID
authn/SPNEGO(already in the file), by adding the attribute
p:activationCondition-refwith a value of
The following examples assume that the condition property
customObject is set to the bean shibboleth.HttpServletRequest.
The following example shows a full-featured script including logging. To see the log lines in the IdP's log file, you need to configure a logger with name "shibboleth.SPNEGO.ActivationCondition" and log level "DEBUG" in your log configuration file conf/logback.xml.
The following code examples might be helpful for you to implement your script.
The following boilerplate code might be helpful for you to test your scripts outside of the IdP. (The script creates a mock custom object that can be used to test various functionality.)
While implementing your activation condition, you may encounter the following problems:
- The script produces a runtime error while it is executed. In this case, you will find an error message in the IdP's log file (i.e. logs/idp-process.log) telling more about the error. The log line will look similar to this:
2015-11-13 11:44:13,408 - ERROR [net.shibboleth.idp.profile.logic.ScriptedPredicate:119] - Anonymous Scripted Predicate : Error while executing Predicate script
- The script runs successfully, but evaluates to false. This means that the authn/SPNEGO flow won't be available.
If you added logging to your script and adapted your logging configuration accordingly, as described above, you will find appropriate log messages in the IdP's log file (i.e. logs/idp-process.log).
Auto Login Management
Users can enable or disable the auto-login cookie by visiting a URL at the path "/idp/profile/user/prefs". You can customize the page displayed by modifying the view template in views/user-prefs.vm.
An alternative is to include the authn/SPNEGO login flow as an "extended" login flow within the authn/Password login flow and allow the users to enable auto-login on the Password login page. This is described below.
Extending the Password Login Flow
If you use the SPNEGO flow in "opt-in" mode, you may want to allow users to login with SPNEGO as an alternative to a Password login. You can make the SPNEGO login option available on the Password login page, and optionally allow users to enable auto-login with SPNEGO at the same time.
Note that SPNEGO as an alternative login method to the Password login is available only if the relying party doesn't request a specific authentication context class that is incompatible with the use of SPNEGO (which would typically be a sign of a poor SP configuration as well). If SPNEGO is not compatible with the requested authentication context class, SPNEGO won't be available on the Password login page. This is checked for you by the default login form template as part of the extended flow feature.
The Extended Flow feature of the Password flow is not specific to the use of SPNEGO and is described more generally in the PasswordAuthnConfiguration / Extended Flow Calling topic. SPNEGO-specific examples are included below.
First, you need to configure SPNEGO as an extended login flow in conf/authn/password-authn-config.xml. Add "SPNEGO" to the bean called shibboleth.authn.Password.ExtendedFlows, and add the auto-login signaling parameter to the bean called shibboleth.authn.Password.ExtendedFlowParameters:
Then, you (usually) will want to adapt the view template view/login.vm. The default view template creates a login button for each extended flow, but doesn't include on option to enable "auto-login". You may adapt the default view by replacing the existing "#foreach" loop by the following code, or you may want to replace it more fully to provide a more appropriate presentation model for the use of SPNEGO in your environment.
The sequence of behavior during the login process is normally as follows:
- First, the SPNEGO login flow will be tried. If the user hasn't enabled auto-login, SPNEGO authentication won't run. If the user has enabled auto-login, SPNEGO authentication is run. If it succeeds, the authentication step has finished. If it fails, the Password login flow will be run next (if it's configured as available flow).
- If SPNEGO didn't successfully run before, the Password login flow will be run and the login page will be shown.
- The user can choose to login with SPNEGO. Optionally, the user can enable the "auto-login" option (if enabled in the view template). If the user has chosen SPNEGO, the SPNEGO login flow will run. This time, the flow ignores the auto-login setting and SPNEGO authentication is run. If it succeeds, the authentication process is typically finished. The "auto-login" cookie is set if the user enabled that option on the Password login page. If SPNEGO fails, the user will be returned to the Password login page.
SPNEGO User Interface
If Internet Explorer is used and Kerberos is not available, and no activation condition (see above) is configured or doesn't prevent this from happening, the browser may show a login dialog box to the user instead of an error page. Users may be confused and may not be able to recover from this situation, so this case should be avoided by configuring an appropriate activation condition if possible.
The view rendering the error page is named views/spnego-unavailable.vm and can be adapted to your needs.
As the SPNEGO authentication doesn't result in a simple username (e.g. "jdoe") but in a Kerberos Principal name that includes the Kerberos realm (e.g. "jdoe@EXAMPLE.ORG"), your current configuration of the attribute resolution might not work, because it expects a simple username to look up the user in the user directory.
You basically have two options to extend your configuration:
- Configure the Subject Canonicalization to transform the Kerberos Principal name into a simple username. This method works if the username that is contained in the Kerberos Principal name is identical to the regular username. Then, you don't need to adapt the existing attribute resolution configuration.
- Adapt the configuration of the attribute resolution to additionally support Kerberos Principal names besides simple usernames. Choose this approach if the subject canonicalization doesn't lead to unique usernames (e.g. because you have multiple realms and the usernames are not unique over all realms; or because this canonicalization may affect e-mail addresses used as usernames, too.)
The following two subsections describe both approaches.
Configuring Subject Canonicalization
This approach requires that each Kerberos Principal name can be transformed to a unique simple username.
You can configure a transformation in the file conf/c14n/simple-subject-c14n-config.xml. See Subject Canonicalization for further information about the available options.
For example, the following transformation rule removes the realm "@EXAMPLE.ORG":
Note: In case you allow your users to use the e-mail address for username/password login, this rule may also apply to such an e-mail address in case the user entered the address in upper case. If this may affect you, please have a look at the next approach.
Adapting the Attribute Resolver Configuration
In this approach, the username is kept as is, but the attribute resolver defines additional attributes from the Kerberos Principal name that can be used in the user directory lookup filter template.
The following example defines the attributes
krb_principalname and krb_domain that are used in the user directory lookup filter template:
Then, the user directory lookup filter template needs to be adapted to use the attributes defined above:
See Attribute Resolver for a detailed description of the attribute resolution configuration.
Testing the Availability of SPNEGO in the Browser
If you have extended the Password login flow to include SPNEGO as an alternative login method, as described above, you may further optimize the password login page to show the SPNEGO login only if it's really available in the browser by using AJAX technology. Although this is fully optional, it may lead to a better user experience.
This section doesn't explain in detail how to do this. The real implementation is left to you as part of the customization of the login page template. But we provide an example of a JSP script that checks whether the browser is capable of providing a Kerberos ticket.
The following JSP script checks whether the browser is capable of providing a Kerberos ticket. You need to install this script to edit-webapp/spnego-test.jsp and rebuild the WAR file and restart the application container to make it available. This example script initiates a SPNEGO login and returns a status in JSON. You can adapt this scripts to your needs. The login page can send an AJAX query to the URL path /idp/spnego-test.jsp and then evaluate the result.
Note: This example is not part of the main IdP distribution. It's just documented here for interested people.
The beans defined in authn/spnego-authn-config.xml follow:
|shibboleth.authn.SPNEGO.externalAuthnPath||String||Value of idp.authn.spnego.externalAuthnPath property||Servlet-relative path to the SPNEGO external authentication implementation|
|shibboleth.authn.SPNEGO.externalAuthnPathStrategy3.3||Function<ProfileRequestContext,String>||A constant function returning the bean value above||A function that returns the redirection expression to use for the protected resource|
|shibboleth.authn.SPNEGO.EnforceRun||Boolean||False||Whether to always try to run SPNEGO, independent of the user's auto-login setting (i.e., try to run for all users)|
|shibboleth.authn.SPNEGO.Krb5.RefreshConfig||Boolean||False||Whether to reload the underlying Kerberos configuration (generally in /etc/krb5.conf) on every login attempt|
|shibboleth.authn.SPNEGO.Krb5.Realms||Collection<KerberosRealmSettings>||sample||List of service principal names and credentials. At least one is required.|
|shibboleth.KerberosRealmSettings||KerberosRealmSettings||Abstract bean used to define beans of this type|
|shibboleth.authn.SPNEGO.matchExpression3.3||Pattern||Regular expression to match username against|
|shibboleth.authn.SPNEGO.ClassifiedMessageMap||Map<String,List<String>>||various||A map between defined error/warning conditions and events and implementation-specific message fragments to map to them|
|idp.authn.spnego.externalAuthnPath||String||/Authn/SPNEGO||Servlet-relative path to the SPNEGO external authentication implementation|
There is no equivalent V2 feature, but this implementation is based on a V2 extension contribution from the FHNW (see Kerberos Login Handler).