Using Genetic Algorithms for Deadline-Constrained Monitor Selection in Dynamic Computer Networks

In this paper we address the problem of selecting a minimal number of optimally positioned monitors for capturing network traffic in dynamic computer network environments. Requirements of computer network monitoring change frequently, e.g., due to indicators of an ongoing attack, which requires a continuous optimization and adaptation of the monitoring state. The monitor selection problem can be mapped to the vertex cover problem which NP-complete. Therefore, we propose a genetic algorithm (GA) as optimization heuristic for obtaining an appropriate solution in adequate time. In case of incidents in the monitored computer network, it is necessary to adapt the monitoring strategy swiftly, reliably and frequently. This is why we constrain the heuristic using a hard deadline for the optimization process indicated by the number of processed generations of the GA. Necessary parameters for the GA were studied in order to optimize them with respect to the given time constraint. The GA was applied to different computer network graph models generated using the Barabasi-Albert model with 30 and 100 vertices and the "National Research and Education Network" (NREN) Europe. We show that the proposed GA provides reliable and valuable results in an adequate time.