The Genetic Approach for Design of Virtual Private Networks

Virtual private networks (VPNs) are now practically the only mean of establishing secure data exchange via the public networks (Internet). The problem of designing a virtual private network is to determine the required number of VPN channels, connecting the fragments of the distributed network infrastructure to each other. However, if the number of fragments is great, this problem becomes rather complex. In addition, the specific feature of solving the problem of designing VPN is the necessity for joint consideration of bandwidth, reliability and cost of the network. This paper offers an approach to solve this problem basing on genetic algorithms. On the basis of the queuing theory the paper examines the analytical models of bandwidth, reliability and cost of the network. The resulting analytical expressions for estimating these properties are then used to construct the fitness function of the genetic algorithm. The structure of the chromosome of the genetic algorithm is proved as a set of elements of the VPN connectivity matrix, lying above the main diagonal. Experimental results obtained for various modes of operation of the distributed infrastructure showed that the optimal structures of the VPN, obtained on the basis of the proposed approach, by the generalized indicator of efficiency have advantage up to 40 percent compared to standard options.