From: Alan Ruttenberg <alanr@media.mit.edu>
Date: Mon, 1 Nov 93 16:24:51 PST
To: cypherpunks@toad.com
Subject: anonymity and privacy in email
Message-ID: <9311012346.AA06090@media.mit.edu>
MIME-Version: 1.0
Content-Type: text/plain



I've looked over the instructions for the anonymous remailer and hals'
instructions, and I have a few thoughts concerning the attempted
guarantee of anonymity and privacy.

In all cases, privacy is guaranteed only if you trust the remailer.
I'll take as a given that this is the case.

But suppose that a response is mailed in plaintext using an encrypted
return address method. The privacy of that message can be violated by
someone who had enough power and interest to monitor incoming mail to
the destination site, since mail and message are unencrypted as the
response enters its destination's mail queue. This is not very much
power to have. The sysop the destination can do this, as can a person
at any gateway between the final remailer and you.

That much can be prevented if you trust the originater of the message
and you have them encrypt their reply using your public key. But
suppose that you have a malicious respondant who wishes to expose your
identity, and has a good guess as to where you might be. Then the
responder needs to send a tagged message and their accomplice needs to
monitor incoming mail looking for that tag. The only way around this
that I see is to have a trusted remailer to which you have given a
public key to use when remailing mail addressed to you.

A second problem concerns the time ordering of incoming and outgoing
messages from a remailer. Consider the one remailer case, as I believe
that the argument holds for chained remailings as well. Suppose that
you are able to monitor the incoming and outgoing feeds of the
remailer. Further, you can identify mail which goes to the remailer
(as opposed to other persons at that site) by reading the to: header.
Suppose that you have a method to identify outgoing mail from the
remailer (from some header) such as "From: nobody@alumni.cco.caltech.edu".

If messages are processed by the remailer in a fifo manner, then you
can identify the recipient of any incoming message assuming that you
get synchronized at some point.  One can to get around this, I think,
is by deliberately scrambling the message processing order, and
perhaps inserting enough fake messages that the monitoring agent can
no longer reliably synchronize.

I'm new to the list, and apologize if I'm repeating previous
commentary. 

-alan