Nelson B wrote:
...
Not really.  Without the authentication, any proxy, including the
so-called transparent proxies, could descrypt all traffic in both
directions without the end parties detecting it.


So, we are saying here that, because there is a small
threat of an active/compromised node doing an MITM,
there is no point in protecting against passive
eavesdropping, which is a demonstrably larger threat?

Is this logic sufficient to justify effectively
denying users any protection against eavesdropping,
within the open, non-commercial world, as befits
the open source community?


There are entire countries whose internet access all passes through
transparent proxies, so their governments can snoop.  If they could
do MITM attacks, you can bet they would.

Governments are trying to snoop on what, exactly? Credit cards? Doesn't make a lot of sense.

Mail?  Most web mail systems don't use any form
of HTTPS, whether it be cert'd or not.  Straight
HTTP.  Getting them to use any form of encryption
would be fantastic.  It'd certainly be a lot easier
if they could bootstrap their webmail servers into
a self-signed cert, accepted by the browsers, and
then upgrade to an expensive CA-signed cert if the
traffic warranted it.

People worried about government snooping on email
based comms generally do in fact get found and
beaten up badly.  Some of them get killed for
their efforts.  They then look at various offerings,
which I guess vary in their nature.  OpenPGP is
widely used by the human rights people for example,
and has been since the early 90s, it was one of the
core user groups since forever.  (If anyone is keen
on *serious* threat models, check out cryptorights.org)

In this case, I'd suggest that if the HRC people
were using HTTPS, they should use a cert.  But, as
the browsers that would be used are potentially
compromised, they would have to validate the browser
somehow.  Hard problem.

So, they might end up having to use OpenPGP on
floppies, install onto a machine, run the program
from the floppy, and then use webmail to transmit
the mail. In which case they'd be happy with any
form of encryption because they've already protected
themselves using other means.

Right now, HTTPS is basically limited to merchants
doing, e.g., credit card stuff or similar.  If the
servers and browsers weren't so serious about merchants
and other server operators being charged hard currency
for running what is basically open source software, the
notion that governments are a threat might make more
sense, simply because there would be non-commercial
usage of HTTPS.  I.e., HRC.  But, for now, no, sorry,
HTTPS is too expensive for the average non-profit,
so I don't see governments being interested.  Correct
me if I'm wrong, please!


> They cannot do undetected
MITM on https, today, beceause of cert based authentication.


Almost all traffic is over HTTP.  There is a bit
of ecommerce over HTTPS.  Is this for real?

Unless you are talking about just your common or
garden criminal bureaucrat working for a government
and also doing a bit of credit card snooping on the
side.  Granted, many governments employ / encourage
criminals, but they are still subject to economic
forces and would rather steal 10,000 credit cards
by hacking than sit there hoping a foreigner comes
along into a net cafe.


I spent some time in one of them this past year.  You can bet I
was particularly careful to ensure I had uncompromised software and
an uncompromised root CA list.  It would take only one compromised
root CA for them to be able to do MITM attacks on all https traffic.


If they compromised the browser you were using,
they could then compromise the traffic from that
machine - unless you used a CA cert list.

But, if they could compromise the browser, and/or
its root CA list, they could also compromise the
entire machine?

What was the threat model here?


Oh, and cert based secure AIM was my friend.


Certs are great if they're available and costless.

They're just not costless.  And, the decision
of the browser and the server to insist on their
usage means that it puts a lot of pressure on
things like CACert to reduce their cost, so we
can see the use of this software by the ordinary
people, rather than by the payments people.

Currently, the cost of certs is the primary
reason that HTTPS has achieved less than 1%
penetration of the market over the last decade.

Presumably AIM does it differently, by using
one cert at the server.  But, if it did it
from user-to-user, as per p2p, then one could
be sure that there would be a desire to reduce
those certs down to their natural zero cost.


iang _______________________________________________ mozilla-crypto mailing list [EMAIL PROTECTED] http://mail.mozilla.org/listinfo/mozilla-crypto

Reply via email to