I'd be interested in reading about any potential research in this area, it's quite a fascinating unsolved problem. Like Bitcoin's solution to the P2P untrusted consensus problem (Two Generals Problem), I would think that the solution to ensuring individuality (see the Sybil attack [2]) lies in some approach that isn't 100% failsafe but probabistically ensures the desired result is highly likely (see 51% attack).
I think a web of trust [3] is something that could work. Given a system of nodes with public keys (the most common form of identification in P2P), a possible approach to estimate the 'individuality' of some identity/key would be to measure its degree of relationships with other nodes who vouch for its authenticity, weighted according to the 'individuality' of the other nodes (something that could be solved with an iterative approach, as in PageRank/EigenTrust). The idea being that while a malicious node might vouch for many of its identities, it doesn't have any link to any other node and thus would be weighted lower.
Turns out this exact approach had been developed in 2011 [5] (just found out now, wow, see fig.1 on page 3) and also in other research with SybilGuard [4]. I'm not sure of the limitations of these formalised definitions, but it looks to me like much of P2P research (see PolderCast, a marvellous innovation) -- the possibilities are never realised until someone implements it in software.
[1] https://en.wikipedia.org/wiki/Two_Generals%27_Problem
[2] http://en.wikipedia.org/wiki/Sybil_attack
[3] http://en.wikipedia.org/wiki/Web_of_trust
[4] http://www.math.cmu.edu/~adf/research/SybilGuard.pdf
[5] https://ccl.northwestern.edu/papers/2011/kurve.pdf