There's been some academic work on identifying these attacks recently. Most of the work I've seen crawls social networks directly. You could try contacting the authors to see if they will share their data sets.

Representative example: SybilLimit: A Near-Optimal Social Network Defense against Sybil Attacks Haifeng Yu, Phillip B. Gibbons, Michael Kaminsky,Feng Xiao http://www.comp.nus.edu.sg/~yuhf/yuh-sybillimit.pdf

A more recent paper looks at a data set from Digg to discover vote manipulation; the authors are interested in further applications and would probably appreciate your contact.

Sybil Resilient Content Rating D.M. Tran, B. Min, J. Li, L. Subramanian http://www.cs.nyu.edu/~lakshmi/nsdi09.pdf