Re: Inclusion of the issuer and serial number in authenticated in format

Denis Pinkas writes:
> (Response from Jim)
> > Denis,
 
> > This is not sufficent to fix your problem.  If the Postal Service
> > in Banana Republic is using the same name, why would they not use 
> > the same serial number as well?  
> In the case I described, they do use the same serial number.
I believe that you are mistaking Jim's point. He was suggesting that the
Postal Service use the same serial number in its own (self-signed)
certificate that the Postal Service in the US uses. This produces an
identical sequence either way.

> > Then you would not have any advantage by putting in the
> > sequence to the issuer.
> Not exactly. The next sequence (assuming it ends up to a trust point)
> will allow to make the difference between the Postal Service from the US
> and the postal Service from the Banana Republic. If this is not yet a
> trust point, then you add another leg and so on until you reach a trust
> point. A I said, this is one possibility among others.
But supposing that our cracker is churning out certificates willy-nilly,
he can reproduce and replace as many issuer-serialnumber pairs as he
wishes. The point at which his attack fails is when the chain you're
validating reaches a trust point. The solution to this is that you don't
trust the certificates in the message independant of other verification,
and a sequence of IaSNs doesn't affect this one way or the other.

To recap: Jim's suggestion was a solution to the problem of a cracker
substituting one of my certs for the other, when both contain the same
public key, but they have different legal meanings. I'm not aware of a
CA anywhere which has seperate certs for itself with different dnames
and the same public key; I'd imagine that CA would be asking for
trouble.

> Instead we should have something like:
> CertUID ::=     SEQUENCE {
>       issuerAndserialNumber   IssuerAndserialNumber,
>       certificateHash         CertHash,
> }
> The certUID is the Certificate Unique Identifier.
I have no problem with this as an authenticated attribute.

Andrew.