[90140] in Cypherpunks
"No one complains" / Stylometry
daemon@ATHENA.MIT.EDU (Randall Farmer)
Sat Nov 15 22:36:10 1997
Date: Sat, 15 Nov 1997 21:13:29 -0600 (CST)
From: Randall Farmer <rfarmer@HiWAAY.net>
cc: cypherpunks@toad.com
In-Reply-To: <199711150104.TAA01930@harrier.sasknet.sk.ca>
Reply-To: Randall Farmer <rfarmer@HiWAAY.net>
> No one complains. Apparently everyone with an ounce of moral sense has left
> the list long ago.
Silence doesn't necessarily denote agreement, especially from someone (for
example, me :) who avoids political posts altogether...
============================================================================
That stylometry thing (figuring out who wrote a message from its content)
seems to be a big problem for people using remailers. Initially, I thought you
could avoid it by using some software to tinker with your words. Although that
can confuse some attacks, there are still author-dependent things you can't
hide so easily. For example, someone with a sufficiently big sample of your
writing (i.e., all a nym's public traffic) could measure, say, the number words
per sentence, or how likely you are to use certain punctuation/function words,
and use that to match it to something you wrote under your own name. And the
measurement part isn't rocket science, either -- I slapped together a simple
stylometer from basic UNIX tools (sed, sort, uniq, grep) and a calculator, and
it appears to actually _work_.
(In addition to being able to say two messages are by the same person, it looks
like you can also tell a little bit about the author, but as for how much and
what...?)
-------------------------------------------------------------------------------
Randall Farmer
rfarmer@hiwaay.net
http://hiwaay.net/~rfarmer
