Predicting location of one hop proxy users

If anyone is interested in this stuff, I recommend taking a look at "How Much Anonymity Does Network Latency Leak?", by Hopper, Vasserman, and Chan-Tin, published in ACM Transactions on Information and System Security. (There's a PDF link listed on http://www.freehaven.net/anonbib/, which is a very large and comprehensive list of papers about these kinds of things.)

The second half of the paper deals with an attack very similar to the one described here. They're trying to do it against multi-hop proxies like Tor though, and so what they have to do is a lot more complicated. Some of the multi-hop proxy stuff is on (in my opinion) somewhat shaky ground since it relies on an older published attack that's only been verified to work when there were ~15 Tor routers on the network (now there are ~2500), but other than that it's a very solid paper.

(My background: I've been doing research related to attacks on Tor since my sophomore year, and I'm now writing my senior undergrad thesis on that research.)