[security] more timing

[Dan Lyke]

On Fri, 27 Oct 2006 09:45:59 -0700, Alaric Dailey wrote:
> Yet another argument for encrypting EVERYTHING between the end  
> points, as well as using something verifiable for keys, rather
> than just something self-signed like PGP or SSH.

While I agree that protocols should be implemented such that  
eavesdropping isn't a concern, I've got two particular issues here:

As I've mentioned before, "something verifiable for keys" simply means  
"we trust the Certificate Authority". I ran that phrase past some of  
my friends who've run network services for over a decade in various  
different forms (I started an ISP with one of 'em back in '93, before  
I left networking for computer graphics), and after they stopped  
rolling on the floor laughing they said that they're of the opinion  
that DNS spoofing in the latest version of BIND is way way harder than  
social engineering a CA.

And since I've been involved in getting a (legitimate) certificate  
recently, I tend to agree.

Second, the specific example here involves a compromised machine and a  
rather clumsy attempt at cracking further services. Encrypting what  
happens between that computer and others in those circumstances is  
like locking a screen door.

Dan