Re: Fixed Size Cell

[Roger Dingledine <arma@xxxxxxx>]

On Mon, Feb 15, 2010 at 12:00:27PM +0100, Runa Sandvik wrote:
> On Mon, Feb 15, 2010 at 11:25 AM, Sameer Ali <nasirmtn@xxxxxxxxxxxxxx> wrote:
> > hi all,
> 
> Hello,
> 
> > I am new and start research in the field of anonymous communication. Could
> > someone tell me please, why TOR use fixed size cell (all cells have same
> > length)?
> 
> Your question was answered by Mauricio Pasquier in another thread that
> you started. See
> http://archives.seul.org/or/talk/Feb-2010/msg00064.html

Mauricio's answer is "fixed size cells helps against traffic analysis
because your data is indistinguishable from anyone's else data."

More generally, we use fixed-size cells because we haven't given up on
the hope that somebody will hand us a cool trick that makes end-to-end
correlation attacks meaningfully slower. Right now we assume they're
instantaneous in all circumstances, but maybe that won't be true forever.

(For more background on end-to-end correlation, see this blog post:
https://blog.torproject.org/blog/one-cell-enough )

We could probably switch to variable-sized cells (we even have them in
the spec and use them in a few cases). It's looking less likely that
somebody will figure out how to secure low-latency anonymity systems
against end-to-end correlation. But I haven't totally given up hope. :)

Also, it looks like Tor's cell chunking assists us to some extent against
website fingerprinting attacks:
http://freehaven.net/anonbib/#ccsw09-fingerprinting
So it may be useful to keep it even if we don't expect defense against
end-to-end correlation. Or maybe the latest website fingerprinting
attacks could become much better if they take into account more of Tor's
architecture during the attack. Hard to say.

--Roger

***********************************************************************
To unsubscribe, send an e-mail to majordomo@xxxxxxxxxxxxxx with
unsubscribe or-talk    in the body. http://archives.seul.org/or/talk/