Network tomography using genetic algorithms

Wide Area Network monitoring has become increasingly important to deliver QoS demanded by customers of Skype, Facebook, Twitter etc or to detect/prevent attacks on networks carrying out sensitive tasks e.g. SCADA networks controlling the electrical grid. Unfortunately the task of monitoring paths becomes increasingly prohibitive as the size of the network increases. Fortunately, recent research has devised interesting approaches for Network Tomography using dimensionality reduction to first simplify the scope of the problem by choosing vital network parameters that must be monitored and then applying statistical techniques to make accurate prediction for unmonitored network parameters from monitored; thus making the problem of monitoring large networks scalable. The latter part of network parameter prediction falls in the domain of optimization problems. Recent work on biologically inspired genetic approaches to solving such optimization problems offers much flexibility in finding an optimal solution. We investigate the feasibility on using genetic algorithms for the network tomography problem.