Standard for embedding KeyInfo

[Kelly Graus <ke...@toltech.net>]

I'm working on signing an XML document using a X509 certificate.  As 
part of the signing process, I am appending DSIGKeyInfoX509 information 
in the signature (by calling appendX509Data on the DSIGSignature 
object).  Once that is there, I am manually adding the name of the 
certificate, and then using that name to find the certificate when 
verifying.
Is there a standard as to what types of information should be stored 
there?  And once the data is stored, is there an automated way of 
loading the certificate based on the data?

Kelly

Re: Standard for embedding KeyInfo

[Arshad Noor <ar...@strongauth.com>]

Agree with many of the comments made by Scott.

In the architecture of StrongKey (www.strongkey.org), a symmetric
key-management software, we've created the concept of a Global Key
ID (GKID) to identify every symmetric key used to encrypt something.

We chose to use the XML Encryption schema for encrypted content, &
since XML Encryption uses KeyInfo from the XML Signature namespace,
we had to come up with a "standard" for the KeyName - and GKID was
it.  Now, within the confines of a given Symmetric Key Management
System in an enterprise, an application using the StrongKey API will
know where to get the symmetric key, as well as which key to get:

..
<ds:KeyInfo>
   <ds:KeyName>1-10</ds:KeyName>
   <ds:RetrievalMethod>
       http://localhost:8080/symkeyServlet/getsymkey
   </ds:RetrievalMethod>
</ds:KeyInfo>
..

So, Scott is right.  There are no rules about these things yet,
but we're sort of trying to create some standards for symmetric
key management services through OASIS (see details on EKMI-TC at):

http://www.oasis-open.org/committees/tc_home.php?wg_abbrev=ekmi

Arshad Noor
StrongAuth, Inc.

Scott Cantor wrote:
>>I'm working on signing an XML document using a X509 certificate.  As
>>part of the signing process, I am appending DSIGKeyInfoX509 information
>>in the signature (by calling appendX509Data on the DSIGSignature
>>object).  Once that is there, I am manually adding the name of the
>>certificate, and then using that name to find the certificate when
>>verifying.
> 
> 
> Name meaning...? There aren't really any names that unambiguously work in
> the absence of a specific context. DNs are useless as there is no global
> PKI, so names are always relative to a deployment scenario.
> 
> 
>>Is there a standard as to what types of information should be stored
> 
> there?
> 
> No. Embedding certificates is far and away the most common approach.
> 
> 
>>And once the data is stored, is there an automated way of
>>loading the certificate based on the data?
> 
> 
> Most libraries provide some kind of key resolution mechanism. This one
> includes a KeyResolver abstraction that returns a key based on a KeyInfo
> object, and supports a couple of basic types when the key is inside the XML.
> This is useless for real applications, since it just verifies the signature
> with a key that is self-evident, but doesn't authenticate the message.
> 
> My OpenSAML project includes a ton of additional code around resolving
> KeyInfo material into credentials and applying trust mechanisms. It's
> extremely complex territory and there are no specs to follow. Shortcuts and
> laziness abounds.
> 
> -- Scott
> 
> 

RE: Standard for embedding KeyInfo

[Scott Cantor <ca...@osu.edu>]

> By this do you mean embedding the certificate in the signature?  If so,
> this sounds like it would work best for us.  Do you have any sample code
> that does this?

I wrap KeyInfo myself now, I don't use this library directly for that part,
it's not as powerful as my code. My old code used it, you can look at:

http://tinyurl.com/2pzlpp

Look at the sign() method.

-- Scott

RE: Standard for embedding KeyInfo

[Scott Cantor <ca...@osu.edu>]

> I'm working on signing an XML document using a X509 certificate.  As
> part of the signing process, I am appending DSIGKeyInfoX509 information
> in the signature (by calling appendX509Data on the DSIGSignature
> object).  Once that is there, I am manually adding the name of the
> certificate, and then using that name to find the certificate when
> verifying.

Name meaning...? There aren't really any names that unambiguously work in
the absence of a specific context. DNs are useless as there is no global
PKI, so names are always relative to a deployment scenario.

> Is there a standard as to what types of information should be stored
there?

No. Embedding certificates is far and away the most common approach.

> And once the data is stored, is there an automated way of
> loading the certificate based on the data?

Most libraries provide some kind of key resolution mechanism. This one
includes a KeyResolver abstraction that returns a key based on a KeyInfo
object, and supports a couple of basic types when the key is inside the XML.
This is useless for real applications, since it just verifies the signature
with a key that is self-evident, but doesn't authenticate the message.

My OpenSAML project includes a ton of additional code around resolving
KeyInfo material into credentials and applying trust mechanisms. It's
extremely complex territory and there are no specs to follow. Shortcuts and
laziness abounds.

-- Scott