[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

More on OCSP and CSL



Jan Meijer wrote:

> I still have some doubts however regarding the feasibility of implementing
> this thing.  It seems to me you open your CA to a nice little DoS attack if
> you do not carefully authorize the certificate suspension.  How to take care
> of the certificate suspension (signature of course, problem solved).

You are right when you say a DoS attack is possible, but the problem, to me,
it is internal to every organization. Let me explain. How CSL is generated
(got from the 24/h phone service, passwd authenticate service... ) this is
something you should solve within the organization and the software you are
using, it is not a matter of standards...

Some time ago we have talked about this problems, one solution would have been
sending an encrypted mail when the certificate is issued with a 'suspension/
revokation' code/passwd (indeed the true path is more complex, anyway lets cut
off details), but this is only ONE solution... :-D

> Second possible problem, how to trust the CSL.  I can trust the CSL because
> it has been signed by the CA on the offline certification server (sorry, had
> to use the term ;).  I cannot ever trust the CSL because it cannot have been
> generated on the offline certification server (because of arguments given
> before).  I don't see a solution to this.

The CSL could be signed by a certificate allowed to sign CSLs. This could be the
same OCSP server: when a certificate is to be moved from one state to another a
new CSL is issued and signed. This could be useful in OCSP implementations as the
OCSP responder can keep informations in the couple CSL/CRL. When a certificate is
found to be suspended, as suggested by many people, a response of 'revoke' can be
given by the OCSP responder with reason extention set to certificateHold...
 
> Third possible problem, how do you recognize the true owner of a certificate
> if he wants to turn his status back to "OK".  Someone explicitly (and
> probably with client authentication or something similar) stated his
> certificate has possibly been compromised.  You will need a good mechanism
> to make sure that the certificate owner is telling the truth when stating
> his certificate has not been compromised.  This might be more of a
> procedural thing, still is important to think about.  Once a certificate has
> been marked as possibly compromised, how to trust it again?
> 
> Jan

Turning his certificate back to OK could be a feature, supported or not by the
organization, requiring more than one pass to be accomplished or not. I think
this is a procedural matter within the organization...

Many (and different) aknoledgemnt methods can be used ( phone, e-mail, fax,
meeting in person) can be adopted to check the ID to be the authorized one.

C'you,

	Massimiliano Pala (madwolf@openca.org)

Attachment: smime.p7s
Description: S/MIME Cryptographic Signature