OpenSAML user discussion

[Brent Putman <>]

Pantvaidya, Vishwajit wrote:
>
> [Pantvaidya, Vishwajit] Brent, thanks a lot for the insight. You mentioned 
> earlier that secret keys are also used to sign SAML responses.



Well, they *can* be, via HMAC signing algorithms.  Most people I think
tend to instead still use public key crypto for XML signatures for the
common use cases, for the same reasons that public key crypto was
developed: solving the key exchange problem.



Signing with a symmetric key via HMAC just mixes bits of the shared
symmetric key material into the data that is digested, so that when the
signature is validated the recipient has to mix in the same key bits in
order to successfully validate it.  AFAIK, it doesn't even use the
associated symmetric cipher algorithm.  That is, the key isn't use as a
"key" per se, but rather just "shared secret bits".



>  Is that also done in a similar way i.e. use a randomly generated secret 
> key to sign and then encrypt the secret key using a public key?
>
>   


No, when you sign with a private key via asymmetric crypto, there is no
symmetric key involved, at least for the algorithms I'm familiar with. 
The amount of data that is encrypted to produce a signature is usually
small, the size of the output of a hash algorithm, e.g. 128 or 160 bits,
or maybe up to 256 or 512 bits with some of the newer SHA variants. 
That's small enough to be encrypted as a single block by even a 512 or
1024 bit RSA key, for example.

--Brent