At 12/4/01 10:02 PM, Eric Smith wrote:

>But my point is that this [encryption without authentication] gains
>nothing.

Well, I guess we disagree, then. I think it gains something: it stops the 
guy in the next cubicle, or the guy in the colo facility, from stealing 
my credit card number or passwords by sniffing Ethernet packets, which I 
would guess is perhaps a more likely attack than someone setting up a 
man-in-the-middle site hijacking.


>If I connect to foobar-random-losers.com that uses a self-signed
>certificate, I have no expectation of any securit or privacy.  I might
>be connecting to people who have hijacked the domain.

[stuff snipped]

>On the internet, on the other hand, there are *trivial* attacks to
>divert or eavesdrop on traffic that can be mounted from halfway
>around the planet.  Sending valuable data over a non-SSL link, or an
>SSL link where the other party uses a self-signed certificate that
>you can't verify, is just asking for trouble.

Well, if we're assuming someone can intercept and alter packets, I gave 
an example in my previous message of a situation where someone could 
trivially hijack an "authenticated" domain, too. He simply obtains a 
certificate for a reasonable-looking URL, then uses a man-in-the-middle 
attack to change the non-secure version of a site so that compromised 
"secure" links appear. For example, he intercepts traffic for 
http://www.tigertech.com/ and replaces the https://www.tigertech.com/ 
secure ordering link with one pointing to (for example) 
https://software-payments.com/, which he owns and has obtained a 
certificate for. Unless the customer knows by some external means that 
the secure link shouldn't actually go to https://software-payments.com/, 
authentication does no good whatsoever.

I often see secure ordering links that take you to a different "store" 
URL that seems completely unrelated (other than page design), and with 
which I'm unfamiliar; any of those could be hijackings, and I doubt many 
of us sophisticates think twice about it. Realistically, if you came 
across a link on my site that said you can order my software through 
https://software-payments.com/, and found a reasonable-looking secure 
page when you got there, I doubt you'd think twice -- you'd have no way 
of knowing my non-secure site links had been hijacked.

As I said, I completely agree that authentication *is* useful in many 
cases, and does add a lot of security if you can externally verify that 
the information the CA attested to (company's physical address, for 
example) is correct. But for some other uses, people are going to accept 
the certificates blindly without verifying (or even being able to verify) 
that they're accurate. In these cases, why should I have to pay for 
(expensive) authentication just so I can use (free) encryption?

Anyway, this is off topic and I'll shut up. I guess my on-topic comment 
is that like everyone else, I'm completely in favor of cheaper 
certificates that are authenticated only by the domain's admin contact, 
so it seems there's no disagreement between any of us as to the relevant 
issue  :-)

--
Robert L Mathews, Tiger Technologies

Reply via email to