'A tattoo is not a face. Ethical aspects of tattoo-based biometrics' by Fabio Bacchini and Ludovica Lorusso in (2018) 16(2)
Journal of Information, Communication and Ethics in Society 110-122 comments
This study aims to explore the ethical and social issues of tattoo recognition technology (TRT) and tattoo similarity detection technology (TSDT), which are expected to be increasingly used by state and local police departments and law enforcement agencies.
The authors argue
Biometrics is the science that aims at measuring and analysing a person’s unique characteristics, both physical and behavioural, for identification and verification purposes. Biometric technologies are used for identification or recognition to determine who the person is, through one-to-many comparison, and for verification or authentication to determine whether the person is who she claims to be, through one-to-one comparison (Mordini and Petrini, 2007). This comparison is made between a “live” digital image of a piece of the body (face, eyes and fingers) and an image of the same piece of the body previously recorded and archived into databases. This previously recorded body information can be retrieved all the time we aim to identify one person or verify her identity. Among the wide-ranging applications of biometric technologies for people verification and identification are the improvement of security against criminal acts and terrorist attacks and forensic uses such as the identification of the perpetrator of an offence. Traditional biometric technologies for people verification and identification include face recognition, gait recognition, fingerprint recognition and iris recognition (Li and Jain, 2011; Jain, Flynn and Ross, 2008; Tistarelli et al., 2009). Recently, so-called soft biometric traits, such as scars, marks and tattoos, are being increasingly used to complement primary biometric identification systems (Lee et al., 2008; Heflin et al., 2012).
Tattoo recognition technology (TRT) is one of the emerging fields in biometrics. In fact, the spread of tattooing in Western countries since the 1990s has been particularly impressive. Many studies report that the incidence of tattooing among respondents in North America and Europe is approximately 20 to 25 per cent (Laumann and Derick, 2006; Swami et al., 2016; Tate and Shelton, 2008); as this percentage is rapidly increasing, it is easy to understand why biometrics is looking at tattoos as a new frontier for identifying people – or at least for integrating their identification. Recurring to tattoo recognition is particularly useful in criminal identification, for instance, when a primary biometric trait such as face or fingerprint is not available.
Of course, not all tattoos are visible (all the time); Dillingh et al. (2016) calculate that 29 per cent of the tattooed population has at least one “subjectively visible tattoo” (i.e. a tattoo that is reported as “normally visible to everyone” by the subject), and 12 per cent has at least one “objectively visible tattoo” (i.e. a tattoo placed on the face, head, neck or hands – meaning that the tattoo is visible even when the individual is wearing a suit and not just when she/he wears a T-shirt and/or shorts).
Most TRT systems developed so far use a keyword-based tattoo image matching (e.g. the ANSI/NIST-ITL1-2011 standard), which is actually highly inefficient due to limitation of the available vocabulary, frequent need of multiple keywords to label a single tattoo image and subjectivity of classification. Recently, alternative retrieval systems were created that extract key-points from tattoo images and use an unsupervised ensemble ranking algorithm to measure the visual similarity between two tattoo images (e.g. tattoo-ID system developed by Jain et al., 2007, 2009; Lee et al., 2008, 2012; Manger, 2012). Some scholars are even starting to develop automatic tattoo sketch to image matching methods, which are particularly useful in cases in which the tattoo image of a suspect is not available, but a sketch of the tattoo can be drawn based on the verbal description provided by an eyewitness or the victim (Han and Jain, 2013).
In this paper, we explore the ethical and social issues of TRT. We want to understand whether TRT deserves special ethical attention in comparison with face recognition and other standard biometric technologies. Although in biometric literature it is customary to introduce tattoo-based biometrics as just another application of biometrics that apparently does not exhibit new specific ethical worries, we believe that the nature of tattoos is such that the opposite presupposition should be advanced before inquiry. In fact, tattoos are meaningful objects, and are capable to reveal much information about us, our identities and our beliefs and affiliations. We will show that tattoos are even more ethically sensitive than faces and that tattoos’ uses for criminal identification or criminal prevention should be carefully monitored in advance, in all their ethical aspects.
Along with TRT, we also examine tattoo similarity detection technology (TSDT), which is aimed at identifying similarity-based classes of tattoos and consequently classes of tattooed individuals supposed to share similar psychological, ideological, cultural, social, religious as well as behavioural properties. After reporting how in particular FBI and the National Institute of Standards and Technology in USA are involved in developing TSDT, we argue that TSDT raises red flags for advocates of civil rights, for instance, because it threatens, in multiple ways, the liberties protected by the First Amendment to the USA Constitution. We remark that TSDT classifies as very high risk for privacy, civil liberty and civil rights according to the criterion for a biometric technology implementation’s ethical risk degree assessment recently proposed by Garvie et al. (2016); and we claim that TSDT also represents a possible factor negatively affecting the health of the people knowing or suspecting to be subjected to a special surveillance in virtue of their tattoos.
The major ethical concern raised by TSDT is perhaps its relying on premises very similar to those founding racist thoughts. We stress that TSDT presupposes the usefulness of human classification based on visible features for prediction of attitudes and behaviours as well as for criminal prevention – an idea that we can find to ground many forms of racism.
Not only does TSDT constitute the unacceptable negative discrimination of tattooed as opposed to un-tattooed people, and of individuals having visible as opposed to those having invisible tattoos, it exhibits what we call indirect negative discrimination against certain demographic groups that are found more frequently tattooed than others and, consequently, are overrepresented in tattoos databases, thus being affected by disproportionately higher risk to be found as a match for a given suspect. We believe it is important to cast light on these indirect forms of negative discrimination that would otherwise easily remain undetected.
