Efficient network attack graph generation

Attack graphs are often used by penetration testing teams to represent the individual steps an attacker could use to compromise a network. The graphs built manually by these teams though are often incomplete and require substantial effort to create. Because of this, automated network attack graph generation is an area that has enjoyed significant research over the last several years. This dissertation further develops the body of knowledge in automated network attack graphs specifically focused on proving its usefulness to protecting real networks. I show that attack graph construction and analysis can be automated while maintaining characteristics of real networks, that attack graphs can be used to assess the resiliency of candidate enterprise networks and that attack graphs can be scaled to enterprise sized networks.