On Wed, Jan 03, 2007 at 03:54:04PM +0000, Michael Rogers wrote: > Matthew Toseland wrote: > > - Locations of nodes are strongly clustered around 0.0. > > Could this be due to some kind of bias in the collected data? For > example if the network is less of a small world than expected, is it > possible that you're only sampling swap requests within one area of the > network? > > Or a more paranoid explanation: if I regularly reset my location to > (Math.random() * 0.02 - 0.01), swapping would spread these locations > around the network and the nodes would end up clustered around 0.0. > > How hard would it be to carry out this attack? Every node that joins the > network creates a properly-distributed location, and every node that > leaves the network destroys either a properly-distributed or a > badly-distributed location. Depending on the churn rate, a single node > resetting its location every few hours might be enough.
The other possibility is that this is happening naturally somehow. My theory was that the node moves peripheral locations to peripheral nodes, and then the peripheral nodes drop off the network. There is some theoretical basis for this but as usual Oskar is in a coma. > > Is there any way to prevent this? Oskar's suggestion of occasionally > switching to a random location would mitigate the problem, but not > prevent it entirely. > > Cheers, > Michael -------------- next part -------------- A non-text attachment was scrubbed... Name: signature.asc Type: application/pgp-signature Size: 189 bytes Desc: Digital signature URL: <https://emu.freenetproject.org/pipermail/devl/attachments/20070115/7e9308b0/attachment.pgp>
