| dc.description.abstract | Physical security systems are intended to prevent or mitigate potentially catastrophic
loss of property or life. Decisions regarding the selection of one system or
configuration of resources over another may be viewed as design decisions within a
risk theoretic setting. The problem of revealing a clear preference among design alternatives,
using only a partial or inexact delineation of event probabilities, is examined.
In this dissertation, an analytical framework for the assessment of the risk associated
with a physical security system is presented. Linear programming is used
to determine bounds on the expected utility of an alternative, and conditions for
the separation of preferences among alternatives are shown. If distinguishable preferences
do not exist, techniques to determine what information may help to separate
preferences are presented. The linear programming approach leads to identification
of vulnerabilities in a security system through an examination of the solution to the
dual problem.
Security of a hypothetical military forward operating base is considered as an
illustrative example. For two alternative security schemes, the uncertainty inherent in
the scenario is represented using probability assessments consisting of bounds on event
probabilities and exact probability assignments. Application of the framework reveals
no separation of preferences between the alternatives. Examination of the primal and dual solutions to the linear programming problems, however, reveals insights into
information which, if obtained, could lead to a separation of preferences as well as
information on vulnerabilities in one of the alternative security postures. | en |
