FACE-OFF: A SURVEY OF FACIAL RECOGNITION TECHNOLOGY, APPLICATIONS, CONCERNS, AND RECOMMENDATIONS

Abstract

This thesis surveys the field of facial recognition, in which technology can automatically identify an individual from an image or video. First, an explanation of how facial recognition works is provided, noting efficacy, success metrics, limits to growth. Second, applications of this technology, such as alternatives and current uses, are discussed. However, no nascent technology comes without concerns. In particular, facial recognition raises concerns of erosions of privacy, as emphasized in the results of our psychological survey on Mechanical Turk, the replacement of human judgment by machine judgment, dataset vulnerability, and system inaccuracy. Furthermore, a growing network of camera technology extending from the public to the private sphere can enable a surveillance state, thus curtailing civil liberties. Current regulations and strategies to inhibit facial recognition illuminate gaps in contemporary responses to these concerns. This thesis concludes with specific policies to be implemented by the federal government of the United States (in the legislative and executive branches), a self-regulation code for creators of facial recognition and private investors, best practices for individuals to protect their private data, and recommendation for future work.