Optimal Attack Path Generation Based on Supervised Kohonen Neural Network

Abstract

Attack graph is a general paradigm to model the weakness of an information system network and all possible attack sequences that attackers can obtain specific targets. In real systems, a vast majority of attack graph generation methods suffer from the states explosion issue. However, if we can predict which attack actions will own the maximum probability to be exploited by intruders precisely, namely finding the optimal attack path, we can solve this problem. In this paper, we propose an attack graph generation algorithm based on supervised Kohonen neural network. Using this method, we can presage the attack success rate and attack status types which would be attained if attackers successfully exploit vulnerabilities. Based on these results and the network topology, a probabilistic matrix and an optimal atomic attack matrix are proposed by us. Finally, the two matrices can be effectively used to generate the optimal attack path. After modeling the optimal path, the core nodes in the target network can be located, and network administrators can enact a series of effective defense strategies according to them.