THE ROLE OF BIOMETRIC APPLICATIONS IN AIR TRANSPORT SECURITY

As the number of air passengers continues to increase worldwide, so do the security demands and challenges in the air transport industry, particularly in lieu of numerous recent terrorist attacks. One of the essential requirements in fulfilling these needs resides in the accurate and timely identification of passengers and other participants in the air transport flow. Biometrics represents identification of individuals based on their quantifiable biological characteristics in automatic pattern recognition systems and is considered one of the most reliable means for personal identification. It is, therefore, not surprising that biometric technologies have an increasing presence in the air transportation industry. The intent of this review is to familiarize readers with biometric tools aimed at establishing and maintaining high security at airports and during flight, as well as to point out promising emerging biometric applications in the field.


Introduction
Airports represent specific locations, given that they are ports of entry and exit approached with high security, but also ordinary workplaces for airport personnel and ordinary travel zones for frequent travelers. 2Few millions of passengers pass through airports worldwide daily.For example, the world's busiest airport, Hartsfield-Jackson Atlanta International Airport in the US, handled more than 94,000,000 passengers in 2013. 3All passengers accessing airport checkpoints and, subsequently their flights, are required to provide a valid identification document (ID).After checked in with the airline, passengers continue through security check point(s), whose intention is to ensure that passengers do not carry dangerous items (weapons, explosives, etc.) that could pose a risk to the integrity of the airport, aircraft, crew and other passengers.This is traditionally done using luggage and passenger screening machines, bomb-sniffing dogs, surveillance camera monitoring, etc.
Yet, despite these safety measures, airports and airlines have become increasingly popular terrorist targets in the past few decades, which, in addition to human fatalities, injuries, panic and psychological consequences, also cause economic damage, and sudden degradation of transportation systems, bringing to attention the issues of transportation networks reliability, emergency response, pre-attack and post-attack counter-terrorism security policies, evacuation strategies, and air traffic congestion mitigation.Furthermore, given that industry experts project 5.9 bn air travelers by 2030, 4 security challenges in the air transport industry are expected to increase.
The central concept in providing collective security to all individuals at airports, as well as protecting aircrafts and airports from unauthorized passengers, outlaws and terrorists, is identification of individuals taking part in the air transport flow.Yet, traditional means of personal identification do not provide the security level that satisfies the current demands in air transportation.Thus, past terrorist actions have increased security efforts globally and reshaped approaches towards airport and airline security, bringing in a new generation of safety measures.Contemporary security measures in large part rely on biometric human identification -human recognition based on their quantifiable biological characteristics.
The primary goal of this paper is to acquaint the reader with biometric human identification, innovative approaches related to the employment of biometric technologies in air transportation, as well as emerging technologies and future directions in the field.

Biometrics 101
Biometrics represents automatic methods for human recognition based on their quantifiable biological characteristics, which can be anatomical (e.g.fingerprint, face, iris), physiological (e.g.heartbeat, brainwaves) and behavioral (e.g.voice, handwriting, gait) (Figure 1).The success of a biological characteristic as a reliable and accurate identifier has been defined by seven parameters: universality (trait exists in all individuals in a population), uniqueness (trait sufficiently differs among individuals), permeance (trait is sufficiently constant over time), measurability (trait can be collected, digitalized and further processed), performance (recognition accuracy), acceptability (agreement of individuals being recognized to present the trait) and circumvention (ease of trait forgery). 5While none of the biological traits score ideally on all seven criteria, nature and requirements of a specific application will dictate the importance of each factor.For instance, if biometric identification is used to restrict access to air traffic control towers to authorized personnel only, low circumvention will have more weight compared to high acceptability in the process of choosing an appropriate biometric trait for this specific application.

Figure 1. Examples of biometric characteristics: fingerprint, iris, keystroke dynamics, gait, face, ear, hand geometry and signature
Although biometrics is an ancient concept, as demonstrated by 31,000 year old handprints ("signature which cannot be forged") discovered in a French cave, 6 contemporary biometrics represents personal identification in automatic biometric systems, developed following the expansion of digital signal processing technologies in the 1960s.An automated biometric system is a pattern recognition system with computer-controlled capture, processing, storage and matching of biometric traits.During registration into a biometric system, a process known as enrollment, individual's biometric trait (i.e., fingerprint image) is captured and digitalized by a sensor, and further processed to achieve high quality image (Figure 2a).Only reproducible and unique components of a biometric trait (i.e., changes in papillary line patterns, minutiae, in a fingerprint image) are then extracted by a feature extraction algorithm, assigned a numerical value (binary vector) and stored as a template in a database.Biometric systems can perform in two modes: verification and identification.The first entails verification of person's identity -an individual presents themselves to a biometric system via a name, PIN, credit card number, etc., based on which reference template for claimed identity is selected from a database and compared to the presented ("live") biometric trait, known as query (Figure 2b).This one-to-one comparison aims to prevent multiple people to use the same identity.On the contrary, identification process begins with immediate biometric trait presentation, capture, processing, feature extraction and template generation.In such a way, query template is compared with all reference templates in an appropriate database (Figure 2c).This one-to-many comparison prevents a single person to use multiple identities.
Two biometric samples collected from the same person (e.g., voice recording) at different time points cannot be identical, due to differing circumstances during sample acquisition (e.g., landline vs. mobile phone, user's dry throat, wind, background noise), changes in a biometric characteristic of interest (e.g., hoarse voice during a cold, high-pitched voice in excitement) and varying interaction between the user and the sensor (e.g., mouth placement with respect to the phone microphone).As a result, biometric systems make two types of errors in verification mode: 1) False match error (FME), which occurs when an algorithm mistakenly classifies an imposter as an authorized user -for example, an unauthorized user is granted access to a protected transportation facility; 2) False non-match error (FNME), which occurs when an algorithm mistakenly classifies authorized user as an imposter -for instance, a passenger is not matched and has to undergo additional checks.The reduction of either of the errors occurs at an expense of the other, thus setting the threshold level is dependent on the requirements and nature of specific applications -lower FNME results in higher usability, while lower FME is a necessary for high se-curity applications.In the US, Transportation Security Administration (TSA) set qualification rates for FME and FNME to be <1.0%.7

Figure 2. Schematic representation of processes in a generic biometric system
Additional errors can occur in biometric systems in both identification and verification modes.Failure to acquire (FTA) error represents users that are unable to provide a usable biometric sample, either because they do not have a biometric characteristic of interest (i.e., mute people) or because it cannot be measured (i.e., severe laryngitis).Failure to enroll (FTE) error occurs when feature extraction cannot occur from previously successfully acquired characteristic (limitations of the technology used).In order to enhance personal identification in biometric systems, a multimodal approach, characterized by the use of two or more biometric characteristics simultaneously, has been increasingly implemented. 8,9 ometric identification has found a growing niche in air transportation, as its use at airports and airlines across the globe is on the rise.This is not surprising in the context of national and international security, as it is essential for every government to identify individuals at border crossings.

Biometric identification approach to border management at airport
Tightened global security measures coupled with the technology advancements have contributed to a worldwide expansion of biometric passports in 2000s.International Civil Aviation Organization (ICAO) has issued Doc 9303 series on Machine Readable Travel Documents, which provide detailed technical specifications and ISO standards regarding biometric passports. 10Implemented standards ensure interoperability between countries and passport vendors, as well as increased security.Biometric passports contain an electronic microchip coupled to an antenna, enabling contactless communication between the chip and a reading device.This chip stores data from the information page and biometric information.Three types of biometric identification systems are supported -facial recognition, fingerprint and iris. 11Facial recognition is a mandatory feature of all biometric passports and it is based on a full frontal (facial) image, while fingerprint and iris are optional.According to ICAO, out of 101 countries which issue biometric passports, 47 use facial image as the only biometric, 51 have opted to include fingerprint recognition, while none of the countries have chosen to store iris image as a secondary biometric. 12ndividual countries have chosen whether to use either verification, identification or both modules within the electronic border control system.For instance, many airports have opted for automatic security/background checks (identification), while traditional identity check scheme (passport control/im-migration officer visually performs facial comparison between the traveler and the digital face photograph in the biometric passport) is still in place of verification.However, in the past years, automated passport control systems have begun replacing traditional passport control desks at international airports in the US, Canada, UK, Netherlands, Estonia, Australia, New Zealand, Abu Dhabi, Japan, etc. E-gates speed up identification time, although issues can also arise (e.g., dirty optical fingerprint reader surface impairs accurate identification).
Biometric identity solutions can also be geared towards increasing convenience and speed of low-risk airport passengers, such as frequent flyers enrolled in Privium and SmartGate programs at airports in the Netherlands and Australia, respectively.More recently, a watch-size wearable digital identity solution, based on unique electrical signals from the heart (cardiac rhythm), has been introduced. 13Virgin Atlantic announced piloting this device, envisioned to allow the enrolled passengers to bypass long waiting lines at airports, allowing them check-in, dropping off their luggage and boarding the plane simply by scanning their wrist.
Importantly, some airports have implemented biometric solutions as part of an immigration procedure.For example, the United Arab Emirates (UAE) launched a nationwide rollout of the "Iris Expellees Tracking and Border Control System" in 2003, with a goal of minimizing illegal immigrants in the country. 14, 15 When an immigrant is to be expelled from the UAE, their iris is scanned at a deportation center and enrolled into a central database operated by the General Directorate of Abu Dhabi Police, Ministry of Interior. 16Irises of all incoming foreigners with new visas are also scanned upon arrival to one of the eight participating UAE airports and compared against this world's largest iris repository. 17In such a way, expellees attempting reentry into the UAE with a new identity and fraudulent travel documents can be identified in real time, reportedly in less than 2s.In eight years, almost 350,000 deportees to the UAE have been identified. 18As in theoretical accuracy performance tests, evaluation of the UAE's iris system demonstrated a lack of FMR, despite 2 tn cross comparisons, due to 'adaptive' decision making process.
Office of Biometric Identity Management (OBIM) in the US, former US-VISIT, employs biometrics to establish and verify identities of international travelers holding a non-US passport.Travelers applying for a US visa at their home country are enrolled into a program, by providing digital images of ten fingerprints and a facial digital photograph.At that time their biometrics are compared against "watch lists" of known criminals and suspected terrorists. 19Following establishment of eligibility to receive a US visa, eligibility to enter the country is determined upon collecting the same biometric data at an arriving airport, verifying that the person entering the country is the same person to whom the visa was issued.This is the largest fingerprint register in the world.Pilot tests at Hartsfield-Jackson Atlanta International Airport and Detroit Metropolitan Wayne County Airport have been made to also implement the system during exiting the US, in order to identify individuals who have overstayed or have stayed illegally in the country. 20However, these attempts have not resulted in the implementation of the program, mainly due to difficulties in ensuring that the person who presented their biometrics while at exit is actually the person who boarded the plane. 21Biometric entry and exit immigration systems also exist in Australia, Canada, Bulgaria, Czech Republic, Ireland, France, Latvia, Ghana, The Netherlands, New Zealand, Saudi Arabia, Taiwan, and United Kingdom. 22

Biometric identification of individuals on airport "safe lists" and "watch lists"
As certain areas of every airport, such as warehouses, hangers and the airport apron, are more sensitive, it is necessary to prohibit access of general public to such areas and allow access only to authorized personnel, such as baggage handlers, maintenance workers and truck drivers delivering cargo.Los Angeles International Airport (LAX), for example, issues 60,000 badges annually 23 and has moved from the legacy system (badge number) to fingerprinting.Canadian Air Transportation Safety Administration (CATSA) credentialing system, in place at 29 Canadian airports and based on biometrics, streamlines both badging and background checks, as well as personnel privileges.During the latter process, employees with permission to enter sensitive areas are enrolled into a biometric database, the so called "safe list" and each time they enter the area during their workday their identity is verified via a biometric trait.LAX is currently storing worker iris images for future use, as multibiometric approach, joint use of several biometric traits including fingerprint, iris, hand geometry and vein patterns in this case, is to provide the least error-prone identification.With the same goal in mind, multibiometric system based on gait, voice and face recognition has been developed and pilot-tested for employee authentication in Euroairport in Switzerland, as part of the Human Monitoring and Authentication using Biodynamic Indicators and Behavioural Analysis (HUMABIO) scientific project. 24onceptually opposite to "safe lists", recognizing personae non gratae at airports is an imperative.Video surveillance of public airport areas occurs routinely at airports worldwide.Images of individuals from video streams taken by Closed Circuit Television (CCTV) or smart phone cameras can be matched against specific databases in biometric systems.These databases known as "watch lists" contain stored templates of known criminals' and terrorists' facial images.This allows for identification of wanted individuals in a mass, at a distance (few tens of meters) and without their knowledge (covertly).Vendors of such technology claim to have achieved up to twenty simultaneous facial recognitions without delay, as well as up to 1 million identity comparisons per second, resulting in immediate alerts upon occurrence of a positive match. 25et, evaluation of facial recognition technology that took place at the Palm Beach airport 26 at processing "load" of 10,000 images per day, resulted in approximately 50% true positives and an average of three false alarms per hour.False positives, resulting from faces being turned away from the camera, facial expressions, etc., also showed to be high in a test conducted at Boston's Logan Airport in 2002, 27 rendering facial recognition at airports impractical as a stand-alone technology, still requiring human operators.While proponents of implementation of such technologies argue that it represents a proactive crime fighting strategy, opposition warns about rights to privacy (people do not know they are being watched), potentials for abuse (once stolen biometric characteristic cannot be replaced) and risks for total surveillance ("big brother is watching you").
Using human gait as a remote biometric tool during airport surveillance has also been considered as an approach since 2000, when Defense Advanced Research Projects Agency (DARPA) in the US launched HumanID at a Distance program.Gait is a convenient biometric characteristic, given that it is contactless and it can be continuously monitored in real-time at a distance, even with lower-resolution images.Since the early years, much research based on static (i.e., leg length) and dynamic walking parameters (i.e., number of steps per minute), using silhouette-based and model-based approaches, has shown reasonable success of gait to differentiate between individuals. 28, 29 , 30 Further research efforts in gait recognition have focused on improving scalability of the initial results, camera viewpoints, effects of walking speed, carrying objects, shoes and clothing on walking patterns, etc. 31 , 32 , 33 , 34 At the National Physics Laboratory, Southampton University, UK, researchers have developed a biometric tunnel with twelve CCTV cameras which allows for automatic gait recognition based on a 3D model of a person's walk. 35Such technology could help airport authorities identify individuals when walking through a monitored area.Research focusing on detection of concealed load 36 via changes in gait is of particular interest for possible identification of suicide bombers at airports.Researchers at Georgia Tech Research Institute, US have used radar to detect a person walking from ~150m distance wearing a simulated suicide bomb vest. 37, 38 Barki and colleagues have attempted to obtain specific (and changed) walking signatures on a test dataset using an inverse kinematic motion model of lower extremities. 39A commercial system already widely employed by military in Iraq and Afghanistan 40 uses a combination of video tracking/gait recognition software and radar technology.Gait rec-ognition software detects individuals and tracks them based on their walking style features and physical attributes, locating them even after they have been obscured by crowds and other objects. 41Low power radar beams repeatedly "inspect" moving subjects in the examined area of interest; reflected beams are then compared to the database of "normal" and "anomalous" responses, determining whether obtained signature corresponds to a person carrying explosive devices or other concealed weapons.Gait recognition software 42 could play a greater role in this system in the future, as researchers are developing tools for analyzing joint movements of walkers, how they correlate with carrying heavy objects, how they change when a person deposits load on the ground, etc. 43 Others have also researched into using knee joint features for biometric identification during airport surveillance, 44 , 45 although the scanning technology itself (Magnetic Resonance Imaging -MRI) is not sufficiently rapid for this task at the moment.
Presented solutions hold promise in identification of suicide bombers and other terrorists at airports at a safe distance, in a non-invasive, covert manner, before they reach airport perimeters or check points.

Biometric profiling approach to airport security
It has been argued that airport screening systems tend to have a high probability of false alarms, resulting in wasted time, resources and effort of the government security agencies which implemented the technology, airport security personnel and passengers selected to allegedly carry prohibited items detected by such screening devices. 46, 47 Furthermore, screening is a challenging task, given that it requires probing millions of individuals in search of a few ones with malicious intents, a process which, in addition to implemented technology, also demands difficult decision-making by security personnel.Thus, some argue that airports should focus on passenger profiling (or behavioral screening), a risk-based approach for detecting suspicious individuals with hostile thoughts.An Israeli company, WeCU Technologies Ltd., has de- [150] NBP • Žurnal za kriminalistiku i pravo veloped a field operational system which is based on audio or visual stimuli specifically designed for relevant target groups, which are expected to elicit a biometric response (conscious or subconscious psychological and behavioral reaction), which can then be captured by hidden cameras and/or concealed sensors either remotely or in an accidental contact. 48For instance, if a terrorist were to be looking at an airport departures screen and a symbol of his terrorist group appears on the screen, he will exhibit an involuntary reaction, such as increased heart rate, breathing and/or temperature, eye fixation and pupil dilatation, etc.Without interrupting regular flow of airport activities, the company claims to be able to measure up to 14 variables and produce results in approximately 35 seconds, with ~95% success rate in tests. 49Physiological variables baseline is to be measured initially and again repeatedly following the presentation of "provoking" stimuli.Therefore, anxiousness of nervous flyers and otherwise disturbed passengers is not expected to create significant false positives.Also focusing on passenger's intent, rather than what they are carrying, Suspect Detection System's Cogito has been tested in the US and Israel's airports. 50, 51 , 52 The system consists of a booth in which a passenger places their left hand on the sensor, wears earphones and sits in front of the screen, answering a set of questions (for instance, "Are you involved in a terrorist activity?"),while their psycho-physiological reaction (i.e., skin electrical conductivity and temperature, left hand and eye movements) is being measured. 53Designed to catch biometric reactions to specific words, which are compared to passenger's reactions while answering baseline questions, as well as to reactions of a peer group (for example, ten previous passengers), the system is designed to pick up those with criminal and terrorist agendas, rather than generally anxious travelers.After the technology reaches acceptable error rates (some tests indicate 8% FMR and 15% FNMR 54 ), it is intended for selected groups of passengers, such as passengers designated as suspicious by officers using The Screening of Passengers by Observation Techniques (SPOT) or travelers flying on known high-risk flights.Future Attribute Screening Technology (FAST) Project 55 , 56 in the US envisions following physiological (i.e., eye movement and blink rate, pupil dilatation, alterations in body temperature), behavioral (i.e., body movements, breathing patterns) and linguistic (i.e., alterations in voice pitch and speech intonation) biometric cues with an idea of predicting future criminal and terrorist actions.The opponents of behavioral approaches designate them pre-crime (after a 2002 movie "Minority Report"), too invasive on individual privacy and ineffective for terrorists who take tranquilizers to mask nervous reactions. 57nnovative approaches to successful biometric identification with a goal of tightening airport security are continuous.Using a new biometric identifier, eyeball movements when following a stimulus projected on a screen, an Israeli company ID-U Biometrics Ltd., has been field testing their product for detection of ordinary passengers and the ones with hostile agendas. 58, 59 Yet, it is crucial that objective, independent testing and evaluation of biometric system performance is carried out by assessing FNMR, FMR, FTA, FTE and throughput rate.Biometric solutions that pass the evaluation criteria and conform to published standards (if any) can be found on TSA's Qualified Products List (QPL).60

Biometric identification ensuring during-flight security
Passenger and crew flow through the airport finalizes with boarding the plane and clearly, it is essential to also maintain high level of security during flight.Biometric solutions have been developed to create secure cockpits, which would limit the entrance into the cockpit to authorized personnel only.Recent tragic events on Germanwings flight 9525 emphasize the importance of adequately approaching this issue.
One of the possible scenarios, proposed in SAFEE program, is that an access to the cockpit is regulated by a fingerprint or another biometric scanner.In such a way, a biometric characteristic being presented live could be matched against a crew database created for each flight.In addition to this, a camera linked to a video display inside the cockpit would allow the pilot or the co-pilot to inspect whether a crew member is attempting to enter the cockpit alone or if they are in duress.Importantly, data of possible imposters would be stored for later evaluation.Additional proposals include surveillance of passengers by cameras and microphones for possible suspicious behavior.
[152] NBP • Žurnal za kriminalistiku i pravo However, if an intruder were to reach the cockpit and attempt hijacking the plane, continuous or on-demand biometric authentication, rather than conventional, one-time-only biometric approaches, could serve as an anti-spoofing strategy.Researchers have been developing such biometric system, which is based on distinctive alpha and beta brain wave patterns recorded by an electroencephalogram (EEG).61,62 Pilot's EEG would thus be stored in a database during enrollment, and brain wave patterns during flight would continuously be compared to the stored template, verifying the assigned pilot is truly operating the plane. 63In case of a non-match due to the changes in brain wave patterns, the system would block the control panel for further use by a hijacker.Although this emerging technology is promising, additional research, as well as adequate testing and evaluations are yet to be carried out.

Conclusion
Automatic biometric identifications represent a contemporary approach to addressing modern day security demands in air transportation.Yet, the expectation of "one fits all" solution is likely far-fetched.For example, it has been noted that long term use of cytostatic capecitabine may cause loss of usable fingerprints for identification. 64Thus, for such patients, as well as for individuals missing an arm, reference templates cannot be generated in biometric systems based on fingerprints.Generally speaking, it is not very likely that a single approach will be satisfactory for airport security and airline safety needs.As a result, quest for biological traits that could serve as biometric characteristics, is continuous.These can be either biological characteristics with previously unknown biometric potential, or biological traits which have long been known to have biometric value, but could not previously be adequately measured or processed.Yet, routine employment of such technologies is not typically quickly around the corner, given that it is necessary that they undergo standardization and objective evaluations by independent bodies. 65,66 urther, biometric systems are not attack-free.For instance, Galbally and colleagues proposed that an iris image can successfully be reconstructed from binary templates and used to trick iris recognition system. 67To that end, much effort is made towards developing superior sensors and algorithms, multibiometric approaches, as well as anti-attack software for diverse attack scenarios. 68pponents of the technology warn that its use is controversial, not always accepted by the public and, in some countries, not properly regulated by law.It has been argued that a concept in which a government or a private airline company handle and manage individual's most private possessions (their own body parts), which, if stolen, misused or manipulated in some other way cannot be replaced, is non-ethical and violates one's rights to privacy.However, the question is whether the right to individual privacy comes before collective security in the air transportation context.In addition, it should be noted that obtaining a biometric passport and participating in the air transport flow are voluntary activities and not government-mandated.These arguments constitute important factors when considering the principle of proportionality in air transportation biometrics.