I was on a customer call recently and this topic came up. In fact since this topic seems to come up fairly frequently - I thought I would describe the recommended model for doing SSO for Web Apps and then doing Identity Propagation across the Back end web services.

The Image below shows a typical flow:

Here is a more detailed drill down of what happens at each step of the flow (the number in red in the diagram maps to the description below of the behind the scenes processing that happens in the stack).

[1] The Web App is protected with OAM and so the typical SSO scenario is applicable. The Web App URL is protected in OAM. The Web Gate intercepts the request from the Browser to the Web App - if there is an OAM (SSO) token - then the Web Gate validates the OAM token. If there is no SSO token - then the user is directed to the login page - user enters credentials, user is authenticated and OAM token is created for that browser session.

[2] Once the Web Gate validates the OAM token - the token is propagated to the WLS Server where the Web App is running. You need to ensure that you have configured the OAM Identity Asserter in the Weblogic domain. If the OAM Identity Asserter is configured, this will end up creating a JAAS Subject.

Details can be found at:

http://docs.oracle.com/cd/E23943_01/doc.1111/e15478/webgate.htm#CACIAEDJ

[3] The Web Service client (in the Web App) is secured with one of the OWSM SAML Client Policies. If secured in this fashion, the OWSM Agent creates a SAML Token from the JAAS Subject (created in [2] by the OAM Identity Asserter) and injects it into the SOAP message.

Steps for securing a JEE JAX-WS Proxy Client using OWSM Policies are documented at:

http://docs.oracle.com/cd/E23943_01/web.1111/b32511/attaching.htm#BABBHHHC

Note: As shown in the diagram - instead of building a JEE Web App - you can also use WebCenter and build portlets. If you are using WebCenter then you can follow the same architecture. Only the steps for securing WebCenter Portlets with OWSM is different.

http://docs.oracle.com/cd/E23943_01/webcenter.1111/e12405/wcadm_security_wss.htm#CIHEBAHB

[4] The SOA Composite App is secured with OWSM SAML Service policy. OWSM Agent intercepts the incoming SOAP request and validates the SAML token and creates a JAAS Subject.

[5] When the SOA Composite App tries to invoke the OSB Proxy Service, the SOA Composite App "Reference" is secured with OWSM SAML Client Policy. Here again OWSM Agent will create a new SAML Token from the JAAS Subject created in [4] by the OWSM Agent and inject it into the SOAP message.

Steps for securing SOA Composite Apps (Service, Reference, Component) are documented at:

http://docs.oracle.com/cd/E23943_01/web.1111/b32511/attaching.htm#CEGDGIHD

[6] When the request reaches the OSB Proxy Service, the Proxy Service is again secured with the OWSM SAML Token Service Policy. So the same steps are performed as in [4]. The end result is a JAAS Subject.

[7] When OSB needs to invoke the Business App Web Service, it goes through the OSB Business Service. The OSB Business Service is secured with OWSM SAML Client Policy and step [5] is repeated.

Steps for securing OSB Proxy Service and OSB Business Services are document at:

http://docs.oracle.com/cd/E23943_01/admin.1111/e15867/proxy_services.htm#OSBAG1097

[8] Finally when the message reaches the Business App Web Service, this service is protected by OWSM SAML Service policy and step [4] is repeated by the OWSM Agent.

Steps for securing Weblogic Web Services, ADF Web Services, etc are documented at:

http://docs.oracle.com/cd/E23943_01/web.1111/b32511/attaching.htm#CEGCJDIF

In the above description for purposes of brevity - I have not described which OWSM SAML policies one should use; OWSM ships with a number of SAML policies, I briefly described some of the trade-offs involved with the various SAML policies here.

The diagram above and the accompanying description of what is happening in each step of the flow - assumes you are using "SAML SV" or SAML Bearer" based policies without an STS.