"Low-Resource Routing Attacks Against Anonymous Systems"

[James Muir]

Ringo Kamens wrote:
> It's very interesting and from what I can deduce, right. However, this
> isn't new news. What *is* new is that it has been proved using math
> and logic. I know that tor will eventually rise to defend this attack
> as other applications such as freenet (i2p also?) have. Another
> implication to consider of this is what if the same ISP controlled the
> entry and exit nodes?

I just skimmed the paper quickly (I'm going to give it a careful read 
tomorrow), but I didn't see too much math in there ;-)  The impressive 
part of the paper seemed to me to be their experimental results (albeit 
using a private 66-node Tor network installed on PlanetLab).

I agree that the principles underlying the attack do not seem to be 
"new".  It was already known that nodes can submit false statistics 
about their uptime and bandwidth to directory authorities.  And it was 
already known that if you control the entry and exit node on a circuit 
you can link initiators and responders using timing analysis.  But maybe 
the paper has some new things to say about the implication of those facts.

Concerning an ISP controlling both entry and exit nodes:  when Tor 
clients build paths, they avoid choosing two nodes on the same /16 
subnet (see path-spec.txt).  So, it does not seem that this is likely to 
happen.

-James