[ILUG] openvpn best practice and revoking

[Gavin McCullagh]

Hi,

I'm experimenting a little with openvpn.  I guess I want to verify that my
understanding is correct now and see what people tend to do.

It seems that the basic authentication process is:

 - create a CA
 - create a key pair for each user, sign it with the CA and distribute the
   private key to them
 - openvpn checks if the connecting key is signed by our CA.
 - openvpn (optionally) checks if the connecting key is not listed as a
   revoked key
 - some optional additional authentication may take place

I had initially assumed that the public key needed to be on the server to
verify the private key, but it seems like that's not important -- instead
the key being signed is what's important.  I had initially imagined that by
deleting the public key off the server, you could disable the private key.
Instead, you revoke the key and a list of revoked keys is checked.

This means that one cannot easily enumerate the working keys in existence
-- you have to keep a list.  So, it's possible for someone to get a key
signed and for you not have any way to know that it exists.  This is a bit
uncomfortable.

Is my understanding correct?

Following on from that, what do people tend to do as best practice?

 - tightly restrict access to the CA key? (presumably!)
 - always put passphrases on created keys?
 - keep that list of signed keys very carefully because it can't be
   verified?
 - use a second authentication mechanism so that a signed key isn't enough?
 - set keys to have relatively short lifetimes (I think the easy-rsa
   scripts set 10 years by default)
 - something else?

Many thanks in advance for any suggestions,

Gavin