Evaluating the effectiveness of sequential aviation security screening policies

Passenger and baggage screening is an essential component of aviation security systems designed to detect and remove threats endangering the safety of air transportation. Upon entering the security checkpoint, passengers are assigned to a multilevel security class system defined by sets of screening devices. The sequential passenger assignment problem has been formulated as a stochastic process, with the objective of maximizing the overall true alarm rate, subject to device capacity constraints. This article introduces three metrics for evaluating the performance of sequential passenger assignment policies with respect to the retrospective optimal solution. The concepts of under-screening and over-screening passengers are defined, from which expressions for the expectation and variance of the number of under-screened passengers are obtained in part through a Markov chain. A conditional probability inequality is used to develop an upper bound on attaining the set of optimal assignments for a given realization of passenger risk. Estimators for the performance metrics are presented for the efficient simulation of cases involving a large number of passenger assignments. The key result is that for populations containing a majority of low-risk passengers, an initial underestimation of the overall risk level produces fewer under-screened passengers in comparison to that which results if the true risk level lies below what was anticipated. Furthermore, fewer passengers are incorrectly assigned to undergo the appropriate screening intensity if security class capacities are biased toward screening the majority of low-risk passengers using a combination of low-intensity detection devices, while reserving the high-intensity, time-consuming devices for the limited number of high-risk passengers.