[2820] in cryptography@c2.net mail archive
Pipe-net redoux
daemon@ATHENA.MIT.EDU (Lucky Green)
Sun Jun 21 11:57:57 1998
Date: Sun, 21 Jun 1998 10:06:14 +0200 (CEST)
From: Lucky Green <shamrock@cypherpunks.to>
To: cypherpunks@algebra.com
cc: cryptography@c2.net
There has been a long standing argument if practical applications of
Chaumian mixes (remailers, TCP anonymizers, etc.) require artificial cover
traffic or not.
In his original Pipe-net post, Wei Dai argued that for chained TCP
anonymizers to provide resistance to traffic analysis, the anonymizers
would have to use constant bandwidth "pipes" amongst themselves. In a
Pipe-net, all nodes send a constant data stream, consisting of either
cover traffic, real data, or a mix thereof. I later suggested to somewhat
alleviate the bandwidth burning requirements of Pipe-net by modulating the
width of the pipe according to long term daily network usage patterns or
some other suitable envelope function.
Others held that there was no need for (semi-)constant cover traffic and
that correlations would get lost in the noise.
I just discovered a paper by the Onion Router folks at NRL that clearly
shows, using multi colored graphs of actual Onion Router usage patterns,
that a non-Pipe-net solution allows for easy traffic analysis.
As to their proposed solution to the problem, it is basically a Pipe-net
where the link padding envelope is modulated with a sine wave, as first
approximation to a function that would more closely match daily network
usage patterns...
http://www.onion-router.net/Vis.html
-- Lucky Green <shamrock@cypherpunks.to> PGP v5 encrypted email preferred.
"Tonga? Where the hell is Tonga? They have Cypherpunks there?"
