Sybil detection in P2P and social networks
We look at the well known graph partitioning problem in the context of distributed simulations. How do you partition the network model under simulation between the hosts? The added complexity of dynamic node and edge weights necessitates looking at this problem from a fresh paradigm. Moreover can we design a distributed algorithm for partitioning the graph? Game Theory gives interesting insights into the design of such distributed algorithms under competing processors. Does a Nash equilibrium exist for such a game? Is it stable?
Crowdsourcing leverages the idea of using a large pool of workers available via the Internet to solve large tasks. The challenge here is to infer correct answers from the set of available answers provided by the users. The use of some probabilistic models renders the inference mechanisms computationally onerous. We propose an alternative model: the "Spammer Hammer and Reverse-Hammer (SHRH)" model that allows simple iterative inference scheme using game theoretic design of node payoff. We also propose another objective based scheme that allows for simple closed form solution to the optimization of a global functtion representing the aggregate confidence level of the inferred answers.