AttributePostLoginC14NConfiguration
- Scott Cantor
Current File(s): conf/c14n/subject-c14n.properties
Format: Properties, Native Spring
Overview
The c14n/attribute post-login subject canonicalization flow extracts a username from an IdPAttribute resolved (primarily) using the attribute resolution function within the IdP. A common use for this flow is to perform a directory or database lookup to map information derived from the authentication process (e.g. the subject of a certificate) into a different value for use within the IdP.
A secondary feature allows for a Java Subject that contains one or more IdPAttributePrincipal objects to be processed directly for an IdPAttribute to pull the value from. This is primarily of use with various "external" authentication options such as SAML proxying, allowing a SAML Attribute decoded from another IdP to be directly consumed and used as a canonical principal name without the hassle of the attribute resolution process (and configuration). This feature can be leveraged by adjusting various properties (see below) to disable use of the Attribute Resolver and reference the Java Subject directly (note this is not referring to the SAML <Subject> element but to the Java object created as a result of all successful authentication flows in the IdP).
General Configuration
Use conf/c14n/subject-c14n.properties to configure this flow, along with the AttributeResolverConfiguration.
If your system is upgraded, you may continue to use conf/c14n/attribute-sourced-subject-c14n-config.xml as before, or you may remove it, while ensuring the new properties are being loaded.
There are two ways this flow can locate a suitable IdPAttribute to use:
By running the "full" Attribute Resolver service (which has some special considerations noted below).
By pulling an IdPAttribute directly from an IdPAttributePrincipal in the input Subject (as mentioned above, this is normally useful when proxying authentication to another IdP). The “Subject” in this case does not refer directly to the Subject of a SAML assertion (e.g., during proxying), but to the Java “object” that represents the result of all authentications internally.
These methods can be combined, in the sense that the list of attributes to search for may be found in either way, so it's possible to run the resolver conditionally and/or check both the Subject and the resolution results. In most cases this is an either/or situation and the resolver won't be used if you expect the data to be in the Subject already.
When pulling directly, you will typically just supply a list of attributes to check for (first value wins), and set the idp.c14n.attribute.resolutionCondition property to "shibboleth.Conditions.FALSE", to turn off the full attribute resolution step, and set the idp.c14n.attribute.resolveFromSubject property to true.
When using the resolver, typically you will supply a list of attributes to resolve and a list of attributes to search for in the results. The first such attribute with a suitable value will supply the username to return.
By default, the only transform applied to the result is a trim of leading or trailing whitespace. Case-folding and regular expression replacements can be added, per the reference section below. The regular expression replacement feature is the only one remaining that still requires XML and you may define that bean, if needed, in conf/c14n/subject-c14n.xml
Using the Attribute Resolver
When relying on the Attribute Resolver with this feature, there are additional considerations of which to be aware.
Usually, the attribute resolver relies on the canonical principal name to do its work (the value of the $resolutionContext.principal variable in scripts or search templates). By definition that isn't possible here (this is the process that provides that value later), but the value of that variable in the resolver can be set by means of a function bean (e.g., using a script) named shibboleth.c14n.attribute.PrincipalNameLookupStrategy. In most cases this function would peek inside the Java Subject being canonicalized and pull out custom bits of information.
The Java Subject this flow is operating on can be accessed via an expression of the form:
profileContext.ensureSubcontext("net.shibboleth.idp.authn.context.SubjectCanonicalizationContext").getSubject()While it would not be a typical case to need to access a value entered by a user except in one specific case noted below, that value will be present inside the Subject as a principal object of type net.shibboleth.idp.authn.principal.UsernamePrincipal. In most other cases, it's simpler to just leave such a value as the canonical principal name and adjust your resolver configuration to accomodate whatever that might be than to try and use the resolver ahead of time to turn it into some other value.
Examples
MFA Example
One situation where this feature is frequently useful:
you need to use a Duo or other 2FA authentication plugin that would rely on the username entered by the user to send to the 2FA service
you want to allow the user to enter their username as one of several different identifiers when they do the initial password authentication
In that case, the same user could end up having multiple different identifiers with the 2FA service, because whichever one they choose to enter at login would be the one sent. This can be solved by using this attribute-based canonicalization, so that you ensure the same user identifier is sent to the 2FA service every time.
The following example illustrates how one could do this, where one has chosen to allow the user to enter either their uid or their e-mail address as the username. Changes need to be made to all of the following, and sample config highlights the changes to each that would enable such. (One would also need to configure your authn handler to allow for either input, such as changing idp.authn.LDAP.userFilter to '(|(uid={user})(mail={user}))'.)
conf/attribute-resolver.xml
conf/c14n/subject-c14n.xml
conf/c14n/subject-c14n.properties
The example takes the user's input name and supplies it to the attribute resolver to do a search of LDAP against the two possible candidate attributes and returns the uid attribute from the directory so that either input maps to a fixed output. There are other ways to configure the resolver that might be more aligned to other searches you need to perform, cache results, etc. but this illustrates the idea.
attribute-resolver.xml
<!-- Example LDAP Connector -->
<resolver:DataConnector id="myLDAP" xsi:type="dc:LDAPDirectory" exportAttributes="uid"
ldapURL="ldap://localhost:10389"
baseDN="ou=People,dc=example,dc=edu"
principal="cn=admin,dc=example,dc=edu"
principalCredential="password">
<dc:FilterTemplate>
<![CDATA[
(|(uid=$resolutionContext.principal)(mail=$resolutionContext.principal))
]]>
</dc:FilterTemplate>
<dc:ReturnAttributes>uid</dc:ReturnAttributes>
</resolver:DataConnector>subject-c14n.xml
subject-c14n.properties
subject-c14n.xml
Reference