Exploring large-scale, distributed system behavior with a focus on information assurance

Large distributed, heterogeneous systems (LDHS's) are increasingly critical to our computational infrastructure, but security research in this area has been difficult. Traditional approaches involve time-consuming and non-repeatable red-team attacks, where results are often incomplete and difficult to interpret. An alternative involves formal analysis of models, but this approach is not feasible with dynamic, heterogeneous systems. The Agent-Based Configurable (ABC) Testbed combines modeling and simulation with innovative visualization. In the ABC Testbed, researchers can model system components at suitable levels of fidelity, execute those models, and analyze the resulting data set, which captures all actions performed by system components, as well as the causes of these actions, or events. The ABC Testbed's visualization tool represents these events as a directed graph, which permits powerful analysis, exploration, and drill-down capabilities. To date, we have developed a prototype of the ABC Testbed, and demonstrated its power by emulating a live experiment and exploring its detailed behavior.