Authentication – Using Specific Claims Provider Trust for Application Group

Well, since nobody did answer, i took the time, setup a test lab and sniffed the HTTPS traffic. Here are my research results in case anybody else will ever come across this thing:

• I still got no official source for this questions
• First off: yes, transitive trust is possible and there is no way to block it except of legal affairs. A propper SLA is the basis for any federation scenario anyways.
• There is no special "setting" to disable or enable it, but using the claims rule engine a trusted partner may configure ANY kind of outgoing claims IF ANY kind of incomming claims are detected, so there is no way to "protect" illegal access.
• In my tests, none of the rule templates that come with ADFS changed the OriginalIssuer property of the claims. One may think: Okay so i can use that property to verify the original source of any claim and build a filter to only allow claims comming directly from (and not traversed through, which by default only affects the Issuer but not the OriginalIssuer property) a trusted partner. But thats wrong, why? Look at the next line.
• As i said, the default templates don't touch the OriginalIssuer property. But you may create custom rules using the rule engine. Using that one, you can change pretty much every claim type, value and their properties. And yes: also the OriginalIssuer. So to the federation provider it looks as the claims are comming directy from the trusted partner, where in reality they're only transformed there.

So, what i would suggest to atleast minimize "transitive" scenarios if they're unwanted, is to check for the OriginalIssuer. It dosn't protect from transitive logins, but a admin would have to explicitly configure it - which would make legal affiars much more easier in a case of SLA voilation. Also, i'm not thinking of the possibility to change the OriginalIssuer as a "bug", in fact: even without that feature, every thrid party software could always make it able to act as a proxy between backend systems and the trusted identity provider. For example the IdP could create shadow accounts for partner (C) - so there will always be a workaround since when using federation, you're giving away the control on who is able to delegate access rights to specific ressources.

Anyway - if you have been as curious as i about how ADFS 2.0 handles transitive trusts, now you know without the need to build a testlab and sniff HTTPS traffic.

A bit late but hope it helps someone. The NameID has a role to play here. Did you see this quote in the "Using AD FS 2.0 as the SAML 2.0 Identity Provider for the Service Provider Sample" readme step 3?

"Note:

· A NameIdentifier claim is not included in the outgoing claim from AD FS by default. This can be added as a Claim transformation rule. This is necessary for logout to perform correctly. "

Was the partner IDP sending a NameID and was your ADFS configured to send a NameID to the RP App? I have configured your very same setup where my IDP and RP STS are both ADFS and this works fine.

Please note the WIF SAML extensions CTP is not supported at this stage. I am assuming this is not a production setup?

Windows Server 2012 AD FS 2.1 is not required for this to work. Although it has some nice extra features coming in the 8.1 server release that might influence your upgrade decisions :)