Mutine: A Mutable Virtual Network Embedding with Game-Theoretic Stochastic Routing

In network virtualization, virtual network embedding is mostly static, which maps each virtual link onto a single predictable path, thus offering a significant advantage for adversaries to eavesdrop or intercept a certain virtual network. However, existing works on multipath embedding just focus on performance and survivability, instead of maximizing the routing unpredictability to avoid link attacks. In this paper, we present a mutable virtual network embedding framework which maps each virtual link onto a set of substrate links with a game-theoretic optimal stochastic routing policy. Firstly, we model the virtual network embedding in the context of stochastic routing with its effectiveness quantified by game theory. Then, in node mapping algorithm, we define a security capacity matrix to evaluate substrate nodes, thus overcoming two disadvantages of existing resource capacity metric. In link mapping algorithm, we work out the optimal stochastic routing policies with satisfying capacity, delay and cycle-free constraints. The simulation results indicate that our framework can significantly improve the probability that packets are not attacked, with little expense of request acceptance ratio and average routing hops.