Early detection of crises using radio traffic analysis
Robert H. Warren
rhwarren@uwaterloo.ca
March 22nd 2003
No matter how much proactive planning is done Emergency Preparedness is
fundamentally about reacting to unexpected situations. The delay between
the triggering event and the notification to the responsible agencies puts
the staff at a disadvantage as they attempt to catch up to a situation.
We propose to research the feasibility of a real-time system capable of
providing an early warning of a developing crisis. This would be done by
tracking the rate at which organizations use their local and national
radio networks. Specifically, we concentrate on the decision algorithms
required for such an application. The intent is to provide a warning order
at an earlier stage than is possible through normal channels, and to do so
in an automated, real-time and objective fashion.
In most situations a triggering event occurs which precipitates a crisis.
As its occurrence is noticed by the local authorities and the scope of its
consequences recognized, additional resources will be incrementally
involved. However, this evolutionary process is not ideal for several
reasons: (1) the information and expertise needed for sensible decisions
may be initially unavailable, (2) requests for additional help or
resources may get miscommunicated, and (3) response lag will occur as each
new agency involved makes its own preparations and assessments.
Much has been written about the need to get the right information to the
right person in a crisis. Even in an ideal situation, managing field
communications is a major logistical problem. In an emergency the problem
is compounded by stress-related mishaps and bad decisions made with
miscommunicated information.
Organizations try to mitigate these errors by establishing communication
channels, call-lists, procedures, and Memoranda of Understanding (MOU)
with other organizations. Such measures have their own limitations as
information can only be collected, analyzed and distributed up to a
certain rate.
To reduce the lag in the communication and assessment process, we propose
to measure the rate of radio network traffic to provide a real-time
warning of a crisis. Fire, rescue and law enforcement organizations are
large users of such networks, as are maintenance, delivery, plant
operations and private security firms.
The relationship between heavy telecommunication usage and a crisis is
well-known: during the events of September 11, 2001, cellular call volume
increased by 50% in the North-Eastern United States (versus a 30%
increase on Mother's day) and wireless Internet traffic increased by 60%
in New York City. 1
Certain agencies have agreements that allow for the monitoring of
one-another's radio networks for items of interest, which requires a
trained communications watch. While anything beyond such agreements raises
some legal and ethical issues, these do not necessarily apply to the
traffic analysis proposed here.
This approach is functional while avoiding a number of privacy and
regulatory issues as the analysis of the utilization rate does not require
the reception of the signal per se. The contents of the messages are not
analyzed, only the rate at which transmissions are made. This coupled with
the available information on ownership of radio channels can be combined
into an activity index for each organization within a certain general
area. 2 By
correlating these indexes for different services, communication trends can
be used to provide advance warnings of upcoming mobilizations.
That Fire Company B has been dispatched to a fire and that all of its
frequencies are extremely busy is of limited interest. That activity at a
local petrochemical plant has spiked and is subsequently followed by fire
and ambulance traffic is of greater interest. It is reasonable to then
assume that further help may be required, for example, to evacuate nearby
homes. It may not be so in the end, but in the meantime a warning notice
can be sent out to other appropriate agencies.
If all radio channels for the local police force are near-saturated at a
location, something of importance is occurring there. Conversely,
complete silence on all police radio nets is an indicator of serious
trouble.
This analysis process can be automated and conveys quality information. It
is immediate and sufficiently detailed for a decision to recall personnel
and begin planning to be made.
Because the project relies only on transmissions detection and not
reception, its radio component is technically simple and the signals can
be detected at a greater range. Developments in Digital Signal Processing
in the last few years have made it possible to monitor large segments of
the radio spectrum in real-time using mostly commercial off the shelf
hardware.
The actual location of antennas, their type and specific coverage are
issues that would be best decided during deployment. We expect that a
small, low-maintenance, low-cost network of distributed nodes would
provide effective results, even if opportunistically co-located.
What we propose to research is the information flow from the signal
detector to the system user. The transmission "hits" from the radio
networks will include modulation, frequency channel and possibly
power/bearing information. The multiple radio transmission sources must be
matched to their owner and then run through a decision algorithm to
determine the stakeholders and the magnitude of the event.
Several historical and geographical databases are available to develop
such decision algorithms. The actual data utilized for this analysis will
be obtained from the Canadian disaster database, radio logs recorded by
different emergency response organizations and if necessary field trials.
3
4 Likewise,
there exists a large body of literature on conventional traffic analysis,
artificial intelligence, distributed databases and electromagnetic
propagation. By using an amalgam of all of these methods, we believe that
key relationships can be discovered which allow for the routing of the
alert to the appropriate persons.
The scope of this research will limit itself to the information processing
aspects of utilizing existing radio telecommunications networks to provide
an early warning of crises. During the 1998 ice storm, cellular phone
traffic increased by as much as 1000% with some
carriers.5
While the nature of the crisis was in itself obvious, the method
described above could possibly have been used to provide an earlier and
reliable indicator of localized crisis.
"Early detection of crises using radio traffic analysis"
Early detection of crises using radio traffic analysis
This document was written by Robert Warren (rhwarren@uwaterloo.ca) on March 22nd 2003. Copyright © 2003-2006 All rights reserved.
Footnotes
- ... City.1
- National Research Council, The Internet under
Crisis Condition, The National Academies Press: Washington, D.C., 2003,
pp. 37-28
- ...
area.2
- Interestingly, this also allow us to tap into the
collective intelligence of bystanders, such as cell-phone users.
- ... trials.3
- http://www.epc-pcc.gc.ca/disaster/cddb_e.pdf
- ...
4
- Preliminary queries with persons involved with emergency
communications indicate that a sizable amount of information is available
on past events in a number of different formats. Whenever possible,
information already in digital format would be preferable.
- ...
carriers.5
- http://www.wirelessweek.com/index.asp?layout=story&articleId=CA4548&stt=001
Rob Warren
School of Computer Science
University of Waterloo
2006-01-31