cacert-sysadm AT lists.cacert.org
Subject: CAcert System Admins discussion list
List archive
- From: IanG <iang AT cacert.org>
- To: Kim Holburn <kim AT holburn.net>
- Cc: CAcert System Administrators <cacert-sysadm AT lists.cacert.org>
- Subject: Re: [Cacert-sysadm] slashdot heads up
- Date: Sat, 19 Jul 2008 16:59:45 +0200
- List-archive: <http://lists.cacert.org/pipermail/cacert-sysadm>
- List-id: CAcert System Admins discussion list <cacert-sysadm.lists.cacert.org>
Kim Holburn wrote:
What Would It Take To Have Open CA Authorities?
http://ask.slashdot.org/article.pl?sid=08/07/18/1721234
I just certified myself at ca-cert:
http://ask.slashdot.org/comments.pl?sid=618797&cid=24246865
http://ask.slashdot.org/comments.pl?sid=618797&cid=24247955
...
They're revoking a certificate roughly every hour, their CRL is 1.9MB in size and from looking at the serial numbers it seems that lots of certificates are pretty close to each other, which means that a significant percentage of issued certs is getting revoked.
This would indicate that their loose verification is being severely exploited by the bad guys.
Well, to answer that, we have to ask a number of questions about what revocation is.
Is there a defined purpose to revocation? Is revocation a declaration that there is a compromise? Or can I revoke because I am bored with those mathematics? Is there some law out there that says I can only revoke when the moon is at full phase?
I would expect that the CPS should state what the regime of revocation is. I'm not interested in looking right now, but IIRC CAcert's CPS says that CAcert's subscribers can revoke any time they like, and reasons are not mentioned. To counterbalance that, it should be remebered that reliance is tested before the Arbitrator, so there might be some wisdom in revoking as soon as the subscriber knows there is a problem, but that still doesn't imply there must be a problem.
Hence, the above logic might be popular but would lack any foundation. In other words, the guy is relying on the false assumption that a revocation only means compromise.
Now are you completely sure that when you add this CA to your store, you also configure the CRL handling properly? For how often do you schedule download of the CRL? Do you really think it's a good idea to download a 1.9MB CRL every 1 hour (there's no OCSP service for their certs, it seems, at least there's no OCSP URL on their CA certs)?
I suspect that you didn't give a thought to this, as well as the majority of people who install CAcert root certificates in their browser, not suspecting what can of worms from security perspective do they open. They probably don't even know what a CRL is for, not to mention checking the CRL handling settings in their browser after they install CAcert's root x.509.
Although subject to some technogeekical babble-analysis, this is primarily a legal problem. The question is one of relying on a cert. The answers to reliance can be found in the CCA and NRP-DaL. Left as an exercise for the reader :)
Another insightful comment was from Philipp Hallam-Baker (I guess, who last I heard was chief scientist at verisign or something similar). He said:
===========================
ObDisclaimer: Not speaking for my employer here. Yes I work for a commercial CA.
Actually you are way off base here. Mozilla and VeriSign are both members of the W3C Web Security Context working group where one of the things that we have been working on is how to better make use of self signed certificates.
I always enjoy reading articles on Slashdot describing what they imagine the optimum strategy for a large public company is.
Making it easier to use encryption with self-signed certificates is a benefit to a large commercial CA. People who use self-signed certificates today are candidates for an upsell to a public accredited domain validate cert later.
===========================
The above is true and helpful, although I would wonder aloud why it is that Mozilla and Verisign are taking so long to do this, if all are now agreed that self-signed certs are good for CAs [1].
Then he goes on to say:
===========================
The basic problem is that people think that a CA sells encryption, that is wrong, we sell authentication and in the case of Class 3 or EV, accountability. I cannot guarantee that the merchant you buy from is honest, or that they will deliver that plasma TV. But I can ensure that they are likely to face consequences if they do.
If people really want to set up an open CA then read my book, the dotCrime Manifesto, I describe what we were trying to do when we set up the idea of CA services in the first place. I think that setting up an open CA would be a bit like *setting up an open source effort to do people's taxes* for them. But someone might work out a way to make it interesting enough for the participants to have it done well.
============================
The last comment is quite insightful (my highlight). Yes, that's a good metaphor for what CAcert has to achieve: a community where we can work to do security stuff that is of that level of importance.
iang
[1] FTR, here is the fuller version of why CAs love self-signed certs, cerca 2004: http://iang.org/ssl/dr_self_signed.html
- [Cacert-sysadm] slashdot heads up, Kim Holburn, 07/19/2008
- Re: [Cacert-sysadm] slashdot heads up, Guillaume ROMAGNY, 07/19/2008
- Re: [Cacert-sysadm] slashdot heads up, IanG, 07/19/2008
Archive powered by MHonArc 2.6.16.