"Steven M. Bellovin" <[EMAIL PROTECTED]> writes:
>On Fri, 01 Feb 2008 13:29:52 +1300
>[EMAIL PROTECTED] (Peter Gutmann) wrote:
>> Actually it doesn't even require X.509 certs.  TLS-SRP and TLS-PSK
>> provide mutual authentication of client and server without any use of
>> X.509.  The only problem has been getting vendors to support it,
>> several smaller implementations support it, it's in the (still
>> unreleased) OpenSSL 0.99, and the browser vendors don't seem to be
>> interested at all, which is a pity because the mutual auth (the
>> server has to prove possession of the shared secret before the client
>> can connect) would significantly raise the bar for phishing attacks.
>> (Anyone have any clout with Firefox or MS?  Without significant
>> browser support it's hard to get any traction, but the browser
>> vendors are too busy chasing phantoms like EV certs).
>The big issue is prompting the user for a password in a way that no one will
>confuse with a web site doing so.

HCI people have been studying this for quite some time, and there's been a lot
of good work done in this area.  Because of the amount of information, I'll
answer indirectly via a link (warning, it's a partial book draft and is
currently ~140 pages long):


Even without this detailed analysis, one of the Mac browsers (Safari?) already
has a quite distinctive password prompt that rolls down out of the menu bar at
the top.  Sure, you can spoof that if you own the browser, but if malware owns
your browser then you're toast anyway.

>It might have been the right thing, once upon a time, but the horse may be
>too far out of the barn by now to make it worthwhile closing the barn door.

That's the response I got from a browser developer when I talked about this
about a year ago, "Sufficiently sophisticated malware can spoof any piece of
browser UI, so let's just give up and admit that the phishers have won".  At
the moment, after 15-odd years of work, the state of the art for both major
secure-channel protocols is to connect to anything listening on port 22 or 443
and then hand over the user's password in plaintext form (although inside a
secure tunnel, as if that made any difference) [0].  This is only just barely
better than the 1970s-era telnet in that the authenticator is still handed
over in plaintext, but at least you can't capture it with a packet sniffer.
Moving to a challenge-response mechanism (which PSK and SRP aren't really,
it's more a bit-commitment since there's no real challenge or response process
[1]) would at least move the security into the late 1980s.

As a side-note, I was talking to a security person from a large (multi-
national) bank recently and they mentioned that they were slowing down on the
push to move to two-factor auth (real two-factor auth with SecurIDs and the
like, not the gimmicks that US banks are using :-) because the problem isn't
authenticating the user, it's authenticating the server and/or the
transaction, and most two-factor auth tokens can't do that.  As a result
they're not going to commit to sinking much more money into something that
doesn't actually solve the problem.  So mutual client/server auth is something
that's of concern to more than just some geeks on security mailing lists, it's
coming onto the radar of large financial institutions as well.


[0] By "443" I mean HTTP over SSL/TLS, obviously.

[1] Actually this is neither challenge-response nor bit-commitment so in the
    absence of anything better I'll propose "failsafe authentication" because
    the other side doesn't get your authenticator unless they can prove they
    already possess it.  In other words if the authentication process fails,
    it fails safe.

The Cryptography Mailing List
Unsubscribe by sending "unsubscribe cryptography" to [EMAIL PROTECTED]

Reply via email to