Question about verifying signatureValue for SAML2 redirect initiated by Shibboleth SP

[Cantor, Scott]

On 2/7/14, 7:34 AM, "Maassen, Helma" <helma.maassen at atos.net> wrote:

>AFAIK the IDP should be able to verify the Signature, using the Public
>Key we did share.

Why do you believe you need signed requests? Chances are you do not, and
so the best answer is to not waste your time and simply turn signing off.

The nest step is to verify that they have the right key, and that they're
not doing something inappropriate with the certificate that has nothing to
do with the signature.

>So what I tried, is to perform the verification myself, using a small
>Java program (source below).

I'm not going to debug your code, but I can already tell you have no
DEFLATE step, so that right there would break it. If you want to implement
something, either read the standard first, or use the Java OpenSAML code.
But the latter is basically pointless because Shibboleth IdPs already use
that code and work fine so it proves nothing new.

And in fact there is no way to prove anything. If you want to clean-room
something and use that as a data point, fine, but the people on the other
end would have to be willing to trust you, and they clearly don't if
they're claiming an implementation that's known to work with many IdPs is
at fault.

>Eventually;
>I need to find out how to convince the IDP-guys (so how to prove), that
>the Signature value in the request sent is valid for the content from the
>request.

There's no way to prove it. You have to choose between an implementation
that's 8+ years old, and whatever this is. The only reported case of
interop failure is with ADFS, and that has never been independently
reproduced.

-- Scott