pem-dev
[Top] [All Lists]

Re: CRLs and COST-PEM System

1993-06-14 13:34:00
Sead,

Sorry that I have not replied sooner, but I was on vacation last Friday.
Allow me to respond to some of your comments.

1.   All verifications of certificates are performed LOCALLY
at the user station, since all valid certificates (your own,
plus all up to the top of PEM hierarchy - IPRA, plus all  of
your  partners  and  all  those from them up to the top) are
earlier retrieved, verified  and  stored  in  your  local
database.

2.   When  you  initialy register, generate your certificate
and send it to your CA for signature, you will receive  back
from  your  local  CA:  (1) your own signed certificate, (2)
your CA's certificate, and  (3)  all  certificates  at  your
branch,  up to the top of the hierarchy.  In the moment when
you receive them, your user  PEM  agent  will  automatically
verify them and if OK, store them in your local database. 

3.   When  you  receive the PEM letter from someone, it will
contain two certificates: partner's and  his/her  CA's.   If
that is not enough for verification (since these may belong,
completely  or  partially,  to a certificate path outside of
yours), PEM  user  agent  will  automatically  send  special
Ceritificate   request  letter  to  the  first  CA  in  your
partner's path whose certificate is missing.  The reply will
contain the required, plus all certificates up to the top of
the partner's path.  When received, all  these  certificates
will be again verified and stored in your database.

One thing that bothers me about this strategy is how a PEM user agent knows
that the response it receives is from the "first CA in your partner's path
whose certificate is missing" is not from some an intruder. In particular
if that CA's secret key becomes compromised and consequently revoked, you
can't trust a PEM message from it. According to your description, I believe
this case would not be properly handled.

I think both John Lowry and Greg Bailey also describe what seems to me to
be significant problems with your approach.

Regards,

Dan Nessett


<Prev in Thread] Current Thread [Next in Thread>