Search RESCUE:

 

RESCUE Login:
User Name:
Password:

 

 

ITR-RESCUE is part of the California Institute for Telecommunications and Information Technology (Calit2) and its IT infrastructure is provided by Responsphere

November 2005

Information sharing research – an important component of RESCUE

Everyone involved in disaster response knows how important it is to get the right information to the right person at the right time, yet this goal is often elusive.  The objective of the Policy-driven Information Sharing Architecture (PISA) research direction within the RESCUE project is to design, develop, and evaluate a flexible, customizable, dynamic, robust, scalable, policy-driven architecture for information sharing that can be used to ensure that the right information flows to the right person at the right time with minimal manual human intervention.

Under a policy-driven approach to information collection and dissemination, policies determine what, when, and where information is collected.  Policies also determine what can be done with the collected information under what conditions, including whose requests to view the information should be accepted (pull dissemination), who should automatically be sent the information (pull dissemination) and in what manner (customized dissemination), what processes can be applied to the information (e.g., can/must it be integrated with information from another source), and the obligations that result from performing these actions on the information (e.g., logging of access).

The PISA effort faces a number of research challenges, including scalability, user-friendly policy management, user acceptance, flexible policy languages, and robustness under attack and heavy load. These include:

• The policies that govern information sharing may be coarse-grained or very fine-grained, making scalability a challenging goal.
• PISA requires the ability for humans to quickly understand complex policies and modify them to fit the current situation.  This is beyond the current state of the art in security policy management.
• PISA requires a flexible policy language that can describe complex combinations of conditions on situations, on potential information recipients, and obligations.  Conflicts between PISA policies must be identified (in advance, where possible) and gracefully resolved at run time, if not before.  In theory, (some combination of) current proposed policy languages offer the needed capabilities to describe these kinds of complex conditions.
• Humans must be willing to use PISA, i.e., they must view PISA as a trusted, helpful asset rather than as an untrustworthy burden.  There are potential obstacles to user acceptance that have been identified in previous research.
• PISA must be robust under attack and under heavy load; like its namesake, it must adapt to difficult conditions without falling.  For example, it should not be possible to amplify the magnitude of a disaster by attacking PISA’s authorization subsystem.  PISA should not become unavailable under situations of heavy load.

These challenges are enormous, and we cannot solve all of them in the context of the RESCUE project.  In RESCUE, we plan to focus our efforts by concentrating on developing and demonstrating PISA in operation in selected hypothetical disaster scenarios.  We will also validate PISA by demonstrating that it subsumes previous proposals for information sharing architectures suitable for use in disasters, and by using PISA in a drill if possible.

Building evacuation drill provides valuable information

A drill conducted on Sept. 22nd by the Responsphere infrastructure team provided a valuable source of information about evacuation behavior. The drill involved a complete evacuation of the Calit2 building, following a real fire alarm.

The data collection process involved the use of indoor and outdoor cameras and a questionnaire. The questionnaire was custom designed for the Calit2 building by the RESCUE team of sociologists, and the drill represented an opportunity to validate this instrument. The validation is necessary in order to be able to use questionnaires with a similar structure in other buildings on campus that do not have the same high tech instrumentation as the Calit2 building. The questionnaire was administered to evacuees upon their return to the building by a team of sociologists located at the main entrances of the building. Forty-two completed questionnaires were returned by the participants.

To summarize the results, we begin by noting 78.5% of all respondents stated they were working at their desk at that time, while 14.3% were engaged in small group meetings (4 or fewer people). When asked what they did between the time they first became aware of the alarm and the time they began to exit the building, 69% of the subjects said they retrieved personal belongings, while 40% said they saved files or shut down equipment before leaving, and 25% finished work or other tasks in progress. As far as giving or receiving information about the evacuation, 37.5% of respondents said they told others to leave the building, and 34% said they were told by others to leave. The contribution of the emergency personnel (floor wardens) was smaller than that of other evacuees in providing information in this case: only 17% declared they were told by emergency personnel to leave the building, and only 7% asked emergency personnel about the exit route.

Another question we asked our subjects is what they believed was the reason for the building evacuation when they first became aware of the alarm. The answers: 12% thought it was a fire, 40% - false alarm, and 45% - a safety drill. However, these results may be biased as a result of the retrospective nature of the data, and the fact that people realized it was not a real fire since they were allowed to return to the building within 5-10 minutes.

The data collected during the drill represents valuable information about what people do between hearing an alarm and evacuating. We observed behavior and belief heterogeneity among actors, and this should be reflected in evacuation simulations. Also, we observed inaccuracy in people’s reports of what they did, when compared with camera feeds.