The state-level attack on the SSL CA security model

Fri Mar 25 16:30:48 UTC 2011

Thanks

The other point I wanted to make is that not every solution is going to work for
every person. If we can improve the current state of things and make life better
for say another 50% of users, that's better than what we have now.
For example in Firefox 4, I could write an extension (if possible) that intercepts the
certificate acceptance dialog and instead does a web query to see how many of
my friends and also their friends  accepted the same cert and at least allow me
to decide with more information than I am presented now. And you could argue
that this should also apply to certs signed by CAs that are in the trust store of the
web browser too.

Just thinking out loud here.

-----------------------------------------------------------------------------------------------
From: Dorn Hetzel [mailto:dorn at hetzel.org] 
Sent: Friday, March 25, 2011 9:24 AM
To: Akyol, Bora A
Cc: Valdis.Kletnieks at vt.edu; nanog group
Subject: Re: The state-level attack on the SSL CA security model

Not entirely unreasonable.  A button for "friend" and then one for "trusted friend" :)
On Fri, Mar 25, 2011 at 12:19 PM, Akyol, Bora A <bora at pnl.gov> wrote:
One could argue that you could try something like the facebook model (or facebook itself). I can see it coming.
Facebook web of trust app ;-)

-----Original Message-----
From: Valdis.Kletnieks at vt.edu [mailto:Valdis.Kletnieks at vt.edu]
Sent: Friday, March 25, 2011 9:05 AM
To: Akyol, Bora A
Cc: Dobbins, Roland; nanog group
Subject: Re: The state-level attack on the SSL CA security model
On Fri, 25 Mar 2011 08:36:12 PDT, "Akyol, Bora A" said:
> Is it far fetched to supplement the existing system with a reputation
> based  model such as PGP? I apologize if this was discussed before.

That would be great, if you could ensure the following:

1) That Joe Sixpack actually knows enough somebodies who are trustable to sign stuff. (If Joe doesn't know them, then it's not a web of trust, it's just the same old CA).

2) That Joe Sixpack doesn't blindly sign stuff himself (I've had to on occasion scrape unknown signatures off my PGP key on the keyservers, when people I've never heard of before have signed my key "just because somebody they recognized signed it").

The PGP model doesn't work for users who are used to clicking everything they see, whether or not they really should...