SRI Security Research Institute, Edith Cowan University, Perth, Western Australia
Attack models can be used to assess network security. Purely graph based attack representation models (e.g., attack graphs) have a state-space explosion problem. Purely tree-based models (e.g., attack trees) cannot capture the path information explicitly. Moreover, the complex relationship between the host and the vulnerability information in attack models create difficulty in adjusting to changes in the network, which is impractical for modern large and dynamic network systems. To deal with these issues, we propose hierarchical attack representation models (HARMs). The main idea is to use two-layer hierarchy to separate the network topology information (in the upper layer) from the vulnerability information of each host (in the lower layer). We compare the HARMs with existing attack models (including attack graph and attack tree) in model complexity in the phase of construction, evaluation and modification.