For Immediate Release: Contact: Tim Farrell June 15, 2009 212.998.6797 tim.farrell@nyu.edu
Facial recognition technology (FRT) has emerged as an attractive solution to contemporary needs for identification, whether for security, law enforcement, or business. As its use increases, those considering adopting it, as well as policymakers, law enforcement and security professionals, the media, and the general public, will benefit from an in-depth analysis that bridges technical, political, and ethical aspects of this powerful tool.
To this end, Helen Nissenbaum, professor of media, culture, and communication at New York University, and Lucas Introna, professor of organization, technology, and ethics at Lancaster University (UK) have published a report, “Facial Recognition Technology: A Survey of Policy and Implementation Issues,” to serve as a guide for administrators, policy makers, public interest advocates, and academics.
Funding for the report was provided by the U.S. Department of Homeland Security, administered through NYU’s Center for Catastrophe Preparedness and Response.
Facial recognition is a form of biometric identification, like iris scanning and fingerprinting. It is particularly attractive for security and law enforcement because it promises the capacity to identify individuals at a distance requiring neither an individual’s knowledge nor cooperation. The hope is that this feature will enable officials to spot known terrorists among a crowd or in areas of vulnerability such as airport security checkpoints.
In fact, facial recognition systems are used to perform three basic tasks: verification, identification, and watch list. In verification, a system verifies identity claims, usually, for cases in which individuals have a pre-existing relationship with an institution, by comparing a probe image of an individual with the images held in a gallery database of the person he or she claims to be. One can imagine an ATM machine using facial recognition technology in this manner.
Identification represents a more difficult task in that a facial recognition system must compare a probe image to many gallery images in an attempt to identify an individual. Identification is even more difficult when the operator of a system does not know whether an individual in question is in the gallery database or not. Finally, in “watch list” mode, a system attempts to discover whether an individual’s image matches any of a gallery of images of known targets (such as known criminals or terrorists).
In all of these applications, failure can take the form of either a false negative or false positive.
Nissenbaum and Introna report that facial recognition systems have performed well in laboratory settings, particularly with verification tasks, but far less well with identification tasks. Performance degrades considerably in natural settings, especially in uncontrolled settings where lighting is poor. This means, at least for now, that facial recognition technologies are nowhere near ready for the task that distinguishes them from most other biometrics, namely, the ability to pick out a face in the crowd.
Nissenbaum and Introna recommend more publicly funded scenario and operational evaluations in order to better support policymakers in making decisions about the appropriate use of facial recognition technology. They also recommend a combination of multi-biometric systems with human operators with sufficient training and discretion to interpret the output of these systems.
The report concludes with a discussion of key ethical questions surrounding the use of facial recognition systems. These include whether subjects know that their images are being captured for gallery databases, whether they are aware that a system is operational, and whether information is being made available to third parties and under what terms. Organizations also owe it to those who are submitted to facial recognition systems to have well-thought out policies for potential false positives and negatives.
Reporters interested in speaking with Nissenbaum are encouraged to call Tim Farrell, NYU Office of Public Affairs, at 212.998.6797.