Matthew Toseland wrote: >> http://dx.doi.org/10.1073/pnas.012582999 >> http://nlsc.ustc.edu.cn/BJKim/PAPER/PRE_CLUS.PDF > > Freenet 0.7 is predicated on the premise that social networks are a lot > closer to small world than to scale free. If they are scale-free then > the darknet is pretty pointless as it's extremely vulnerable to targeted > assassination.
The model in the second paper generates a scale-free degree distribution if nodes are never deleted from the network, which is what's been observed in social networks where nodes are never deleted, eg citation networks, collaboration between actors, sexual contacts. However, when nodes can leave as well as join, the degree distribution is no longer scale free. I don't yet know whether this means the network is no longer vulnerable to targetted attacks. But in any case, I'd be using those models to generate the 'raw' social network between users, then extracting a more realistic darknet topology from the social network - for example there would probably be a limit on the number of darknet links per node, no matter how many social contacts the user had. Cheers, Michael
