RE: Online Certificate Revocation Protocol

["Carlin Covey" <[e-mail]>]

Mark, as I understand it, when a Relying Party encounters a revoked
or expired signature certificate, the RP is supposed to reject the
signature as invalid.  There has been discussion on this list and elsewhere
of historical verification of signatures, i.e. "was this signature
valid at a certain time in the past?"  This is straightforward
if the signature certificate is still valid.

But if the certificate was revoked, particularly for reason of key
compromise, then historical verification of an associated signature
becomes troublesome.  It is also troublesome to verify a signature
after the certificate has expired.  I believe that the expiration
of a certificate relieves the issuing CA of the responsibility to
maintain revocation status for the certificate.  Of course a Relying
Party can continue to use the public key in the expired certificate,
whether for encryption or signature verification, if the RP is willing
to accept the risk that the private key has not been compromised after
the certificate expired.  Basically the public key is no longer certified,
so the previous obligations of the CA and subject of the certificate
no longer obtain.

If the signature must be validated after the certificate has expired
or been revoked, the signed document should be timestamped while the
signature is still valid (certificate not expired or revoked) by a
timestamping authority whose certificate is still valid. Eventually the
authority's signature will expire, and before that happens the
the timestamped document should be encapsulated in another layer of
timestamping.

As an alternative to repeatedly encapsulating the signed document in
additional layers of timestamping, the original timestamp authority
could hold a "private key destruction ceremony" in an attempt
to demonstrate that its retired key will never be compromised.

Another alternative might be for the timestamping authority to assume
the responsibility for reporting compromises of the expired private key.
If the expired private key were ever compromised, it could sign a
"my private key was compromised" notice either with the expired private key,
or with a private key associated with a non-expired certificate with the
same
subject DN as the expired certificate.


Regards,

Carlin

____________________________

-  Carlin Covey
   Cylink Corporation

-----Original Message-----
From: Scherling, Mark [mailto:[e-mail]]
Sent: Tuesday, June 12, 2001 10:41 AM
To: 'Bob Jueneman'; Scherling, Mark; 'Carlin Covey'
Cc: [e-mail]
Subject: RE: Online Certificate Revocation Protocol


My understanding is that if you revoke a certificate or key that the key is
no longer valid for signing and the user cannot use the key or certificate
to gain access.  It does not mean that anything prior to that point is
invalid.  What happens when the certificate or key expires?  Does that mean
that all documents and transactions signed before are invalid? No you can
still validate the signature.  And what happens 10 years from now, the
public key is still valid for that time.  It should be available (either
through archive or on-line) to check to determine if the status was valid.
If I remember correctly the CRL merely indicates that the certificate was
revoked on a specific date.

Now getting into an area I'm not that familiar with, path validation checks
to see if the certificates are valid.  If a certificate is invalid, it
fails.  This only applies to current transactions.  I don't think we have
considered past or expired keys.  It's somewhat analogous to doing a search
on a deed title or any record retrieval.  There has to be somewhere where a
record of the transaction and the signatures attached.  In the case of a
paper document the signatures are with the document, in an electronic
document, where are the signatures, with the document.

How do you verify them?  In paper, you have to compare with other samples.
If you don't have samples, then you have a problem.  How do you prove that
that is the person's signature and there was intent by that person to sign.
In the case of most legal documents, you have a witness signature, usually a
public notary  or a person representing an entity involved in the
transaction.  So we get back to Lynn's notary and how they could be of value
in proving the signature.  There are no right answers, each organization
will determine the risks and create a process based on perceived risks.  The
financial community will demand far greater security measures than most
other organizations, but I think in the end we will require some sort of
notary function on higher value or risky transactions.


-----Original Message-----
From: Bob Jueneman [mailto:[e-mail]]
Sent: Tuesday, June 12, 2001 10:12 AM
To: [e-mail]
Cc: [e-mail]
Subject: RE: Online Certificate Revocation Protocol


Mark, I still disagree.  If a key was deliberately destroyed, then it is in
a considerably less unknown state than it was before. After all, the
question of unknown duplicates, or even known duplicates (escrow) also
applies while the key is in active use, as well.  If the user was sloppy
about the destruction, then maybe he was sloppy about its protection before
the fact, as well, and that is something no notary is going to be able to
help with.

But I think there may be a more fundamental disconnect.

My point is that few, if any, people are likely to go to all of the trouble
to time stamp and notarize messages that are received, just so they can
prove at some point after the certificate has expired (or been revoked) that
it was valid at the time the document was signed or the signature was first
verified.  Unfortunately, having spent the last 15 or so years of my career
in this field, I see a very disappointing amount of uptake for PKI other
than for immediate access control decisions (including SSL certificates).
And despite a number of brave, (and perhaps increasingly desperate) vendors
of time stamping/notarization systems, I see little indication of any
significant deployment of such systems.

Even if PKI is eventually used and accepted by businesses (and I certainly
hope it will be), most transactions are over and done with at the time they
are first received.  Even those that aren't concluded immediately, such as
back-ordered merchandise, are accepted and held in abeyance.  And the number
of times that the signature validity of such a transaction would be
questioned is virtually nil.  Business simply don't operate on the fictional
"let's do business with complete strangers" model.  Instead, virtually every
business, of whatever type, operates on a cyclical, monthly statement type
of basis, and on the basis of trust between the partners -- a trust which
goes far beyond the simple question of identity.

While I don't disagree that it might be nice to have all of the powerful
timestamping, etc., mechanisms, available in case they just might be needed,
I would argue that they are far from necessarily in the vast majority of
cases -- maybe 999,999 out of a million. And the cost and complexity of
establishing such mechanisms may not be worth it. We simply aren't using
(nor very likely to use) PKI for the type of enduring transactions between
strangers (wills, deeds of trust, etc.) that gives rise to the necessity of
nonrepudiation measures.  Even in those cases where documents are examined
long after they were written, a legal challenge to the signature is
exceptionally rare.

The rest of the time, and this was the essence of my previous comment, it is
sufficient to validate the signature at the time it was received, or shortly
thereafter.  If any question arises and the certificate hasn't expired, then
the relying party can revalidate the certificate at that time and will still
receive a valid answer.

Perhaps I should state my assumption explicitly: The probability that a
given certificate will need to be reverified declines very sharply with time
after the initial verification, at least for most relevant business
transactions (not including wills and trusts, obviously). Time stamping is
therefore only required if one believes that the transaction will have
significant probative value at some uncertain time in the future, after the
certificate has expired.

(True notarization, as opposed to a mere time stamped validity check, is
probably even less necessary. An exception might be for those transactions
(typically involving unsophisticated consumers) where it is necessary to
impress them with the ceremonial importance of their acts and to act as a
true witness for their mental competency, absence of coercion, etc. as well
as justifying the high fees typically involved in such a routine
transaction. Buying or selling a house and taking on a mortgage is about the
only such transaction that people do very often that might benefit from true
notarization, and even then it is quite questionable.  After all, if you
move out of your house and let someone else move in, it is pretty obvious
that you approved of the transaction.  And if you don't pay the mortgage,
the bank will foreclose, regardless of the signature or lack thereof.)

If you assume, as I suggested, that private keys and certificates are taken
out of service well before the end of their validity period and new ones
created, i.e., for most transactions the certificate expiration date will be
a year or more in the future, then this is a safe and reasonable assumption
and timestamping is probably not required.  And this is true regardless of
whether the private key was destroyed or not.

Some correspondents have suggested that providing anything less than the
absolute ultimate in security will prevent PKI from ever getting off the
ground.  While I would hesitate to predict whether or not PKI will ever
succeed, the fundamental problem would seem to be due to spending too much
time and effort to mitigate risks that weren't terribly large to begin with,
or didn't fit a reasonable business model.

And even then, classical PKI only provides information about an entity's
identity. And identity, while certainly useful for auditing terms, is
neither absolutely necessary nor sufficient for the ultimate purpose, which
is to decide exactly what that entity is trusted to do, or why the entity
should be trusted at all.  Granted, given someone's identity I can look up
their privileges in my locally maintained directory,.but in that case why do
I need a certificate at all, especially if the keys are ultimately going to
be protected with a password in any case.

Too much mechanism to mitigate too little risk, and this 10 years after the
first PKI standard was first written.  Too little, and much too late, I am
increasingly afraid.

Bob



>>> "Scherling, Mark" <[e-mail]> 06/12/01 09:18AM >>>

If there is a notary in the context of the transaction then the notary would
be liable for the transaction if the certificate and private key that signed
the document originally was proven to be invalid (i.e. key was assumed
destroyed but copy made and copy signed document).  I think that we can
argue that there was no intent by the owner of the key to sign the document,
however their digital signature is attached to the document signed by the
notary, who did not know that the key was destroyed (no record of key
revocation, certificate is valid, so notary signs).

I really like the idea of a notary function but you still need to revoke the
key if it was destroyed.  A key that was destroyed is in an unknown state
(was the key really destroyed and are there no duplicates?).  So the CA must
revoke the key to place it in a known state.  The public key can still be
used to verify transaction prior to the revocation.  However anything after
revocation should be rejected.  I feel that the security risks associated
with leaving a key in an unknown state are far greater than the problems
associated with revoking a key.


-----Original Message-----
From: [e-mail] [mailto:[e-mail]]
Sent: Monday, June 11, 2001 7:04 PM
To: Carlin Covey
Cc: [e-mail]
Subject: RE: Online Certificate Revocation Protocol





... as per aside ... having somebody sign a document ... and then a
notary validate the signature with the public key, and then the notary
signs a composite document ... consisting of the originally signed
document, the signer's public key, and the current time ... and then log it
to a secure audit trail ..... could be done completely w/o the original
signer's certificate .... since in effect, the notary can perform at least
all the feature/function of a RA & CA as part of their function (in effect
the composite document that the notary signs .... is a kind of
certificate).

It isn't absolutely necessary to know any validity period (from a
certificate) of the original signer's public/private key .... it is just
necessary that the notary validates the information as correct at the time
it was signed/validated (and/or can be later shown to be valid at the time
of the signing).

... have you noticed that the postings to the mailing list seems to have
some sort of lag? I've yet to see my original reply to you made at 3:59
(MDT) ... 3 hrs later (presumably you answered the copy of the reply sent
directly).






"Carlin Covey" <[e-mail]> on 06/11/2001 04:42:12 PM

To:   Lynn Wheeler/CA/FDMS/FDC@FDC
cc:   <[e-mail]>
Subject:  RE: Online Certificate Revocation Protocol


Lynn,

I quite agree that notarizing, with or without secure time, is a more
comprehensive solution.  I simply proposed one-time signature keys as an
example of a situation in which the certificate is expressly intended to be
valid after the private key has been destroyed.  Now whether anyone would
want to use one-time signature keys is another matter ....

Regards,

Carlin

____________________________

-  Carlin Covey
   Cylink Corporation

-----Original Message-----
From: [e-mail] [mailto:[e-mail]]
Sent: Monday, June 11, 2001 3:59 PM
To: Carlin Covey
Cc: Bob Jueneman; [e-mail]
Subject: RE: Online Certificate Revocation Protocol




in many cases, notary can include the idea of (secure) time .... i.e. that
not only can you proove who signed it ... but also when it was signed.

in principle, private keys (whether compromised or not) should not be able
to "pre-date" such a notorized, secure "time" signing.

typical solution is either a secure audit trail .... and/or to encapsulate
the signing inside some other transaction/document which includes a secure
time which is then signed by the notary function. The notary function
(wether audit trail or encapsulated function) can also include the business
function of validating/prooving the original signature (aka the notary
attests to the validity of a specific signature at a specific time).

while a one-time key with non-expiring certificate could meet a subset of
the business requirement .... it is not clear how many business processes
would need just the subset w/o needing the rest of the capability (aka, a
secure audit that establishes the validity of a signature executed at a
specific time would subsume the need for a one-time signature key and also
meet additional normal, day-to-day business requirements .... aka not only
is there the issue of what order a sequence of signatures might have taken
place .... but also what order did signatures take place within the context
of real-world events and sequences ... i.e. time).

If you are going to go to all the trouble of a notary ... dump the stuff
with the one-time private key .... and meet the rest of the business
requirements which includes did the signature verify and at what time did
the signature verify.






"Carlin Covey" <[e-mail]>@mail.imc.org on 06/11/2001 10:00:12 AM

Sent by:  [e-mail]


To:   "Bob Jueneman" <[e-mail]>
cc:   <[e-mail]>
Subject:  RE: Online Certificate Revocation Protocol



[Bob Jueneman]:

Indeed, although some have deprecated the concept of a private key validity
period, it makes a great deal of sense to DELIBERATELY destroy a given
signature key, especially a code or certificate signing key, well before
the
corresponding certificate expires.  From the point of view of the
certificate subscriber, this minimizes his risk by making certain that the
key can NOT be compromised, yet the certificate has not expired or been
revoked, so the certificate will continue to validate properly.

[Carlin Covey]:

I agree with Bob.  It might even be desirable to use "one-time" signature
keys for signing particularly important documents, such as major contracts,
wills, etc.   There might even be a "super non-repudiation" policy
associated with the guaranteed destruction of the signature private key.
This might be implemented via some trusted hardware token that generates
the
keypair, signs the document, destroys the private key, and signs a
notification of private key destruction.  Another possibility is some sort
of trusted "key-destruction notary" service that notarizes the document,
and
then destroys the certified one-time signature key as a matter of policy.

Regards,

Carlin

____________________________

-  Carlin Covey
   Cylink Corporation