This paper presents an agent-based evacuation simulation framework in an airport setting. Air transport services have become a nerve system of our society, therefore, aircrafts and airports represent a central component of todays traffic system. A significant heterogeneity of processes and a high levels of surveillance define the uniqueness of this scenario. Building Evacuation is an important application field for the Agent-Based Modeling (ABM). In recent years, there has been an increasing amount of literature on airport analysis and modeling, but not so many have simulated emergency evacuation using agent-based modeling. In this research three hazardous scenario are modeled: (i) an earthquake, (ii) an accident (fire) and (iii) a terrorist attack (shooting). In all three cases the model takes into account the main actions that an agent can perform during the evacuation: the search of the optimal escape route, the leaderfollower behavior and the research of a missing person. Three types of agents are modeled in the simulation: the aircraft passengers, the security agents (or aircraft crew) and the robbers. The first type has an incomplete knowledge of the area, unlike the security agents who can reach the emergency exit walking the shortest path, then their first task is to help the other agents to reach the closest exit. A cross platform Java-based modeling system called Repast Simphony has been used to develop the ABM. This tool provides decision support in a real emergency scenario in order to study the dynamics of evacuation, helping to improve the infrastructure reliability and its response when a critical situation occurs and also assessing the effect of a certain number of security agents in the area. In order to show the implementation issue the framework has been used to model the Fiumicino Leonardo da Vinci International Airport in Rome and the two decks of an Airbus A380-800, however it can be easily extended to other types of environments.