Optimal Attack And Defence Of Large Scale Networks Using Mean Field Theory

We address the issue of large scale network security. It is known that traditional game theory becomes intractable when considering a large number of players, which is a realistic situation in today's networks where a centralized administration is not available. We propose a new model, based on mean field theory, that allows us to obtain optimal decentralised defence policy for any node in the network and optimal attack policy for an attacker. In this way we settle a promising framework for the development of a mean field game theory of large scale network security. We also present a case study with experimental results.