Saddle policies for secure routing in communication networks

In this paper we formalize routing in communication networks as a game between the designer of the routing algorithm and an attacker that attempts to intercept packets. By computing saddle-point solutions to this game, we obtain stochastic routing policies that are secure in the sense that they utilize multiple paths to minimize the probability of packet interception. We also show that the policies obtained for secure routing can be optimal with respect to flow-maximization and load-balancing. These policies thus have applications beyond secure routing and may proof useful, e.g., in wireless networks, where bandwidth and power is at a premium.