-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 freenetwork at web.de wrote: >>>Matthew Toseland wrote: >>> >>>>The main outstanding issue is how frequently we should do path folding. >>>>If it is too slow, it will take too long to converge. But if it is too >>>>fast, then oskar's routing algorithm won't be able to keep up. Is there >>>>any way to determine an optimal time short of alchemy? >>> >>>=20 >>>If I have understood things correctly, on the new network a node can have >>>both darknet connections and opennet connections; in which case, when data >>>comes from a darknet node, it should forward it hijacking the source as >>>beeing itself, otherwise, it would use the usual algorithm. Being on the >>>border of the darknet (giving the darknet a gateway to the opennet) means >>>not giving away any info on darknet nodes. Being fully inside the darknet >>>means you don't know anything about nodes that you've not been introduced >>>to (and none else besides them should try to connect to you either!). >> >>Correct. > > Then what is the implication of border nodes always resetting Source to > themselves? > > I think that would bring to light that they are border-nodes between the > open and the closed network. > > Analysis could be done because those border nodes often route requests > with an HTL < maxHTL (because the request went some time through the > darknet) although they pretend to be the Source; correlation attacks > (border nodes tend to have a higher correlation > "randomness" by previous darknet routing steps than nodes requesting the > files all by themselves); network harvesting with connection analysis (an > harvested opennet node has X routes to other nodes, analysis would reveal > that this node has X connections to other > nodes; border nodes have X to opennet and Y to darknet, a harvesting > would only find the X links but network analysis would reveal X+Y links -> > border node, possible entry point into the darknet: now either send Those > Guys or disconnect every border node found to > separate the smaller darknet from the well-known opennet)
Good points. For those reasons, I believe path folding should happen more often. I have a scenario which may even favour specialization: A node knows 3 things about the nodes implicated in the transfer of some piece of data: 1. It's "source" (the last node that claimed ownership) 2. The node it is actually receiving the data from 3. The node it is passing the data to Now, for all of these nodes it has specialization information. An example of the specialization graph on that piece of data along the path followed when the data is found, could be something like this: 1 -- <-source of request 2 ----- <-candidate 3 ---- 4 ------ 5 -------- 6 ----------- <-candidate 7 --- 8 ------------- <-candidate 9 -------- 10---------------- <-data found here I think that a node should have a higher propability when deciding to claim ownership of the data whenever it finds itself on a peak of specialization with regard to that piece of data (as indicated by the "candidates"). Now, this scenario assumes plenty of specialization and good distribution of the specific piece of data (that's how it is found on a node highly specialized to it). If the data is fresh and/or inserted with a low HTL (i.e. not well distributed), the graph might show something like this: 1 -- <-source of request 2 ----- <-candidate 3 ---- 4 ------ 5 -------- 6 ----------- <-good candidate 7 ------ 8 ---- <-data found here In this case, node 6 is a very good place to store the data anyway. The node also knows that its in a peak, but also knows that the "source" node also has lower specialization than itself. Shouldn't it keep a copy of the data /and/ claim ownership? Of course, I'm not familiar with current node behaviour in these cases (for instance, I don't know when a node decides to store data currently). Consider this just food for thought. You might also consider possible threats to the network when this behaviour is modified maliciously. As before, please correct any faulty assumptions I have made. Doc - -- Theory asserts that although in theory there should be no difference between theory and practice, in practice there IS difference between theory and practice. Constantine Dokolas Sunsoft Ltd. Mimikos Group Tel: (+30) 210-931.7811 Fax: (+30) 210-932.1603 -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.2 (MingW32) Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org iD8DBQFDM8/ev5NVEkAIb1ERAlEkAKCp4H+hQmaGX+oVEyZdBIf9Q9JtvwCfdNio 8frXk3M0/syWREeMk8gOd6w= =ydyR -----END PGP SIGNATURE-----
