Computational Intelligence - February 2017 - 17

Kenneth Lai, Svetlana N.Yanushkevich,
Vlad P. Shmerko, and Shawn C. Eastwood
Biometric Technology Laboratory, Department of Electrical
and Computer Engineering, University of Calgary, CANADA

Abstract-A biometric-enabled watchlist (or a database of persons of interest) is commonly accepted by national and international security agencies. In particular, the facetrait has demonstrated promising performance in large-scale open-set tasks in
forensics and law enforcement scenarios. However, their application in the security of
mass-transit systems, such as e-borders, is very limited. We evaluate the risks of
unwanted effects using Doddington's metric and various databases. The key contribution of this paper is a novel technique for traveler risk categorization using a biometric-enabled watchlist and an evidence-accumulation paradigm, as well as its
impact on e-border performance.

I. Introduction and Objectives

©istockphoto.com/maxiphoto,
no fly sign-image licensed by ingram publishing

mechanism which enables the identification of individuals of interest is known as a
watchlist. Advanced biometrics bring a new horizon in forensic watchlist technologies [1]. It is well documented in forensics that errors can occur in profiling,
searching, matching, and identification using biometric traits [2], [3]. Our study is
aimed at bridging the gap between the forensic biometric and biometric-enabled watchlist
used in contemporary automated border control (e-borders) for homeland security.
Traveler risk assessment in e-borders has the following main components: authentication and watchlist technologies [4]-[6]. In e-borders the central problem is to
improve the performance, such as throughput, and minimize the risks of the border
crossing of an unwanted traveler [7].
The International Air Transport Association (IATA) and International Civil Aviation Organization (ICAO) predict that the volume of international air travelers
will grow at around 4.1% per year. Passenger numbers are expected to reach
7.3 billion by 2034 [8]. Each of these 7.3 billion travelers should be checked via
watchlists at all phases of the traveler's journey including pre-screening, screening,
airport logistics, border crossing, and flight via distributed mass-surveillance.
Traditionally, traveler risk assessment is based on past and present behavior, and
is aiming at predicting potential future threats. These risks are analyzed, systematized, modeled, and various recommendations for their mitigation have been
developed, in particular in [9]-[11].
In e-borders, the problem "Is this traveler on the watchlist?" [4], [7], [12] is
reformulated as "Assess the risk of the traveler being on the watchlist" [5], [13].
Conceptually, this means that the watchlist check [14] is replaced by the watchlistbased inference of risks. We study such risks in this paper.
This paper investigates the unwanted effects of biometric-based watchlists
for mass applications in e-borders. These effects carry additional weight in
traveler risk assessment. For an evaluation of these risks, we conducted
experiments using different approaches and databases, and a unified Doddington's categorization metric. Besides the formalized background of

February 2017 | Ieee ComputatIonal IntellIgenCe magazIne

http://www.istockphoto.com/maxiphoto

Table of Contents for the Digital Edition of Computational Intelligence - February 2017