Hi Pedro,
I'll let Jim follow-up on the details here and see with you the best way to
move forward.
This said, in the current scenario, the isValid is method is not called (it
was originally thought for the username token profile scenario, so the
string password credential would be available, etc), maybe it should be
changed a bit to be used in this scenario too, so that the pushContext is
not needed, according to what you're writing below.
Can you and Jim please review the current implementation of
ElytronSecurityDomainContextImpl and change it to support this SAML token
authentication scenario too?
Thanks
On Mon, Jun 4, 2018 at 3:31 PM, Pedro Igor Silva <psilva(a)redhat.com> wrote:
In Keycloak integration we have a specific security realm
implementation
that expects a principal previously authenticated by a keycloak adapter
(e.g.: using SAML or OIDC) and builds an authorized identity based on it.
Basically, what this security realm does is populate the authorized
idenitty with information from tokens.
Later we complete authentication in Elytron and set the token as a
credential into the identity. It is worth mention that in Keycloak
integration, the adapter is a Elytron HTTP Authentication Mechanism, so we
don't deal directly with the security domain but with the callback handler.
Regarding ElytronSecurityDomainContextImpl, is method pushContext called
after a call to isValid ? If so, the security domain should be set with the
security identity and you don't even need to keep that ThreadLocal ...
On Thu, May 31, 2018 at 7:03 AM, Darran Lofthouse <
darran.lofthouse(a)redhat.com> wrote:
> Just added Pedro in CC so see if he has any suggestions - this is
> sounding similar to the problems he would have needed to handle when he
> added support for KeyCloak integration using the Elytron APIs.
>
> Although the reported problem we are working on is in the context of
> access to the token it does currently sound that there is a missing
> pre-requisite step of tying the authentication to Elytron to we can
> populate a SecurityIdentity. But this does not sound like the first time
> we have needed to approach this.
>
> Regards,
> Darran Lofthouse.
>
>
> On Thu, 31 May 2018 at 10:54 Jim Ma <ema(a)redhat.com> wrote:
>
>> On 05/31/2018 05:37 PM, Darran Lofthouse wrote:
>>
>> So the validation is within Apache CXF - is there an end result to this
>> validation where you have access to everything you need where we could
>> perform some additional steps?
>>
>>
>> After Apache CXF validation, we can get a LoginContext from CXF's
>> exchange message :
https://github.com/apache/cxf/
>> blob/master/core/src/main/java/org/apache/cxf/security/Login
>> SecurityContext.java
>> Can we do something to convert it to an Elytron authenticated identity
>> ?
>> Or we have to hook/replace something with Elytron in CXF's validation
>> to make this work ?
>>
>>
>>
>> On Thu, 31 May 2018 at 10:34 Jim Ma <ema(a)redhat.com> wrote:
>>
>>> The saml validation is now Apache CXF's SAML functionality. We can't
>>> port the CXF's security to rely on
>>> our Elytron.
>>>
>>> On 05/31/2018 05:07 PM, Darran Lofthouse wrote:
>>>
>>> It sounds to me then that the place to start is within the SAML
>>> validation, this is effectively an authentication step so should be ported
>>> over to an Elytron based authentication - the end result of the
>>> authentication would then be the required SecurityIdentity to propagate
>>> from container to container.
>>>
>>>
>>> On Thu, 31 May 2018 at 03:57 Jim Ma <ema(a)redhat.com> wrote:
>>>
>>>> On 05/30/2018 09:47 PM, Darran Lofthouse wrote:
>>>>
>>>> I am currently gathering together some information regarding how the
>>>> JCA subsystem handles the requirement of populating a Subject for
>>>> propagation into a resource adapter, however there is a general question
>>>> about what is attempting to be achieved here.
>>>>
>>>> Once an EJB is secured using WildFly Elytron the associated identity
>>>> is not accessed as a Subject instead it is accessed a SecurityIdentity
the
>>>> current SecurityIdentity can always be retrieved by calling the current
>>>> SecurityDomain: -
>>>>
>>>>
http://wildfly-security.github.io/wildfly-elytron/1.3.x/api-
>>>> javadoc/org/wildfly/security/auth/server/SecurityDomain.
>>>> html#getCurrent--
>>>>
http://wildfly-security.github.io/wildfly-elytron/1.3.x/api-
>>>> javadoc/org/wildfly/security/auth/server/SecurityDomain.
>>>> html#getCurrentSecurityIdentity--
>>>>
>>>> The SecurityIdentity has some similarity with the Subject in that
>>>> amongst other things it also contains a collection of public credentials
>>>> and a collection of private credentials: -
>>>>
>>>>
http://wildfly-security.github.io/wildfly-elytron/1.3.x/api-
>>>> javadoc/org/wildfly/security/auth/server/SecurityIdentity.
>>>> html#getPublicCredentials--
>>>>
http://wildfly-security.github.io/wildfly-elytron/1.3.x/api-
>>>> javadoc/org/wildfly/security/auth/server/SecurityIdentity.
>>>> html#getPrivateCredentials--
>>>>
>>>> So I think the very first question is has the SecurityIdentity been
>>>> correctly populated with any delegated credentials? If not that is
going
>>>> to be a pre-requisite for any follow on steps regardless.
>>>>
>>>> Then secondly what is it that is making use of this identity? Why
>>>> can't it be ported to make use of the Elytron authentication client
APIs
>>>> which amongst other things provide support for delegation from the
current
>>>> identity.
>>>>
>>>>
>>>> If we need to we can look at a conversion to a Subject but we are only
>>>> doing that where it is really required.
>>>>
>>>>
>>>> We don't have the SecurityIdentity populated, there is only
>>>> principal and subject created by jbossws/CXF's saml validator.
>>>> We need to convert the subject/principal to Elytron's
>>>> SecurityIdentity or something else, then later on EJB subystem with
Elytron
>>>> security can retrieve this authenticated info without check it
>>>> twice. So we'd like to know how can we convert a subject/principal
>>>> to Elytron's SecurityIdentity and let Elytron know this is already
>>>> authenticated and authorized.
>>>>
>>>> Thanks,
>>>> Jim
>>>>
>>>>
>>>>
>>>> Regards,
>>>> Darran Lofthouse.
>>>>
>>>>
>>>> On Wed, 30 May 2018 at 10:27 Alessio Soldano <asoldano(a)redhat.com>
>>>> wrote:
>>>>
>>>>> As suggested by Darran, I'm forwarding the message below to the
list
>>>>> on behalf of Jim.
>>>>> The classes Jim is referring to are at
https://github.com/wildfly/wil
>>>>> dfly/tree/master/webservices/server-integration/src/main/
>>>>> java/org/jboss/as/webservices/security
>>>>>
>>>>>
>>>>> ---------- Forwarded message ----------
>>>>> From: Jim Ma <ema(a)redhat.com>
>>>>> Date: Wed, May 30, 2018 at 9:03 AM
>>>>> Subject: Set an authorized identity to EltyronSecurity Context
>>>>> To: Darran Lofthouse <darran.lofthouse(a)redhat.com>
>>>>> Cc: Alessio Soldano <asoldano(a)redhat.com>
>>>>>
>>>>>
>>>>> Hi Darran,
>>>>>
>>>>> We are helping look at a customer issue which requires propagate the
>>>>> authenticated subject from webservice subsystem to
>>>>>
>>>>> ejb subystem. With old security domain , we can do this with
creating
>>>>> a subject :
>>>>>
>>>>> @Override
>>>>> public void pushSubjectContext(final Subject subject, final
>>>>> Principal principal, final Object credential) {
>>>>> AccessController.doPrivileged(new
PrivilegedAction<Void>() {
>>>>>
>>>>> public Void run() {
>>>>> SecurityContext securityContext =
>>>>> SecurityContextAssociation.getSecurityContext();
>>>>> if (securityContext == null) {
>>>>> securityContext = createSecurityContext(getSecur
>>>>> ityDomain());
>>>>> setSecurityContextOnAssociation(securityContext);
>>>>> }
>>>>> securityContext.getUtil().createSubjectInfo(principal, credential,
>>>>> subject);
>>>>> return null;
>>>>> }
>>>>> });
>>>>> }
>>>>>
>>>>>
>>>>> After Elytron, what is the equivalent thing to do this then ejb
can
>>>>> retrieve this security without check this twice ?
>>>>>
>>>>> Thanks,
>>>>>
>>>>> Jim
>>>>>
>>>>>
>>>>>
>>>>>
>>>>> --
>>>>>
>>>>> Alessio Soldano
>>>>>
>>>>> Associate Manager
>>>>>
>>>>> Red Hat
>>>>>
>>>>> <
https://www.redhat.com>
>>>>> <
https://red.ht/sig>
>>>>>
>>>>
>>>>
>>>
>>