A Scalable Approach to Full Attack Graphs Generation

摘要

Attack graphs are valuable vulnerabilities analysis tools to network defenders and may be classified to two kinds by application. One is the partial attack graphs which illustrate the potential interrelations among the known vulnerabilities just related to the given attack goal in the targeted network, the other is full attack graphs which evaluate the potential interrelations among all the known vulnerabilities in the targeted network. The previous approaches to generating full attack graphs are suffering from two issues. One is the effective modeling language for full attack graphs generation and the other is the scalability to large enterprise network. In this paper, we firstly present a novel conceptual model for full attack graph generation that introduces attack pattern simplifying the process of modeling the attacker. Secondly, a formal modeling language VAML is proposed to describe the various elements in the conceptual model. Thirdly, based on VAML, a scalable approach to generate full attack graphs is put forward. The prototype system CAVS has been tested on an operational network with over 150 hosts. We have explored the system's scalability by evaluating simulated networks with up to one thousand hosts and various topologies. The experimental result shows the approach could be applied to large networks.