[freenet-dev] a paper worth reading

[Matthew Toseland]

On Tuesday 10 June 2008 01:44, Ian Clarke wrote:
> I'd recommend taking a look at this:
> 
>   http://www.cl.cam.ac.uk/TechReports/UCAM-CL-TR-637.pdf
> 
> I haven't finished reading it myself, so I won't comment until I do.

Interesting. Not entirely about scale free networks (which suck, as we all 
know): beta networks are somewhat similar to what we're looking at for a 
darknet (some short links and some random/long links), although probably less 
resilient than a real small world network (p ~= 1/d). Simulations assume a 
scale-free network - clearly the closer our network is to scale-free, the 
worse we will be under most attack models. However it looks like the defences 
attempt to evolve it into something closer to what we'd recognise, and to 
some degree they succeed... Rings don't work because there is insufficient 
clustering within the command structure: cliques work better, with each node 
within the broken-up command node being linked to all the others. Arguably 
cliques are more clustered than a typical small world darknet though. 

I'm not sure what to make of the finding that the bigger the cliques the 
better (in terms of impact on us)... It suggests a two tier network, or a set 
of two tier networks, where everyone in the first tier knows everyone else in 
the first tier, and then each of them knows his local "flock"... Delegation 
avoids over-reliance on hubs, again reducing the scale-free vertex degree 
variability of the network, and seems to help further...

Interesting paper, tends to back up the assumptions we've been making 
mostly... I'm not sure how much it directly applies to those trying to use 
darknet cell networks though...
> 
> Ian.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: not available
Url : http://emu.freenetproject.org/pipermail/devl/attachments/20080613/e9f6bd77/attachment.pgp