Using Binary Particle Swarm Optimization for Minimization Analysis of Large-Scale Network Attack Graphs

The aim of the minimization analysis of network attack graphs (NAGs) is to nd a minimum critical set of exploits so that by preventing them an intruder cannot reach his goal using any attack scenario. This problem is, in fact, a constrained optimization problem. In this paper, a binary particle swarm optimization algorithm, called SwarmNAG, is presented for the minimization analysis of large-scale network attack graphs. A penalty function method with a time-varying penalty coe cient is used to convert the constrained optimization problem into an unconstrained problem. Also, a time-varying velocity clamping, a greedy mutation operator and a local search heuristic are used to improve the overall performance of the algorithm. The performance of the SwarmNAG is compared with that of an approximation algorithm for the minimization analysis of several large-scale network attack graphs. The results of the experiments show that the SwarmNAG outperforms the approximation algorithm and nds a critical set of exploits with less cardinality.