Thus spake Gregory Maxwell (gmaxwell@xxxxxxxxx): > On Fri, Aug 10, 2012 at 10:11 PM, Ted Smith <tedks@xxxxxxxxxx> wrote: > > The obvious problem with this (((this, right here, is the productive > > contribution to discussion this email has: it points out the problem > > with your proposed methodologies))) is that it presumes that these top > > 50 .onion domains comprise the majority of .onion traffic through your > > node. I suspect this is not the case. Yeah, unlike exit nodes, this is definitely not the case. Your node can be chosen as either the guard, the introduction or rendezvous point and you will see traffic correlation for page views at a much greater frequency than expected, if you're not careful about it. > > If I'm right, and most of the .onion traffic through any given node is > > over the "long tail", it won't be possible to get anything useful > > without an automated classifier. Bingo. > It's odd that this thread started with a discussion of some sketchy research > which worked by running an automated spider moving enormous amounts of > traffic. > > So the very thing that inspired the conversation ruins the proposed methodology. > > Tisk tisk. Yeah, a related question is "How much 'Illegal/Questionable' traffic through exits actually *is* law enfocement?" It's not all of it, of course. Might not even be most of it. Unless they have automated crawlers... Still, it is a little surprising they can't trace bitcoin yet, though. Maybe they can. I think my bet is also on Silk Road not surviving in the long run for that reason... It's very interesting to watch, for sure. It's like we're getting an extra season of The Wire, except in a much weirder world that couldn't possibly exist except in some Sci Fi novel. -- Mike Perry
Attachment:
signature.asc
Description: Digital signature
_______________________________________________ tor-talk mailing list tor-talk@xxxxxxxxxxxxxxxxxxxx https://lists.torproject.org/cgi-bin/mailman/listinfo/tor-talk