defeating traffic analysis

>Date: Tue, 19 Oct 93 22:43:46 PDT
>From: hfinney@shell.portal.com (Hal Finney)
>Message-Id: <9310200543.AA22959@jobe.shell.portal.com.shell.portal.com>
>Subject: Re: JUDGMENT PROOFING

>This is exactly what anonymous remailers are for - to defeat traffic
>analysis.  The goal is a situation where you, and everyone else, receives
>and sends out a bunch of encrypted traffic every day.  There will be
>no way a snooper can find out exactly what you are reading, what you
>are writing, or whom you are communicating with.  That's what I call
>private.

To really defeat traffic analysis you also have to remove any information
from volume of traffic.  If I don't know any message routing but I do know
that cme sent 10 MB of traffic and you received 10 MB of traffic and the
other 30 people on the anon subnet dealt with only 100B each, ....

So:  have everyone send and receive exactly the same size message to/from the
same sites at the same time every day.


 - Carl

# From: cme@ellisun.sw.stratus.com (Carl Ellison)
# 
# So:  have everyone send and receive exactly the same size message to/from the
# same sites at the same time every day.

That seems so unfortunate that the "average utilizaton" and the
"peak throughput" [I use quotes because I may have the wrong 
technical buzzwords, but you know what I mean] have to be
one and the same.

I've been trying to think of ways around that.  Am I wasting my time?

I've thought of trying to find probabilistic ways of letting you
vary how much your read/write to the net each day, maintaining
some constant statisics (average amount you read/write, standard
deviation, ...), that allows you one day to read a lot more than
the average, when you need to, without giving away that it was
useful information instead of padding. 

But I'm afraid it still gives away partial bits of information --
it is still likely that the hungriest readers and the most verbose
writers match up.

This becomes more on my mind, as I work on DCNet protocols.

References?  Advice?                    strick