Description

ID:Talks are bonus presentations in the exhibit hall and are not components of the formal agenda produced by the FedID Planning Committee.

The purpose of this brief presentation is to present the use of face recognition (FR) as a powerful tool for identification in public and private use, and to address the mistaken assumptions that FR is subject to exploitation in the US. Within the past few years, face recognition (FR) has become an effective, non-intrusive, mainstream technology. Millions who shunned passwords and other protections on mobile devices and computers have enthusiastically embraced FR. The advantage? Unlike our passports, wallets, and keys, we never forget to take our faces wherever we go. The face has long been used for identification. It is primarily how we recognize one another. Use of the face in passports, driver licenses, and identification cards of all types is nearly universal. Humans are quite adept at recognizing family, friends, and celebrities, yet surprisingly poor at distinguishing strangers. Professional passport examiners and border agents have trouble matching, or excluding, the unfamiliar when presented with good quality picture identification. However, machines excel at this task. In 1996, following the recommendations of the 9/11 Commission, Congress mandated a biometric exit system for international departures from the U.S. After much analysis, and maturation of FR technology, the Department of Homeland Security settled upon Face Recognition as the least intrusive and most cost-effective, solution. The face is already used in all passports and virtually all other travel documents. These images are already available on file and used as part of both the arrival and departure process. FR only makes the processes faster, more accurate, and less expensive. Citizens may opt out, but few do. For those citizens who participate nothing additional is retained after 12 hours. Some have voiced concern over police use of the technology. Claims of algorithm bias have been lodged, citing decade old research while FR was under development and not ready for wide scale implementation. In fact, extensive testing shows higher identity verification performance for dark skinned individuals compared to light skinned individuals, for more than the hundred best-performing algorithms. Fears have been stoked over possible mass surveillance, citing the example of surveillance in China. However, no one in U.S. government or industry has advocated such a system. Existing CCTV cameras cannot readily transition, nor is coverage sufficient for general identification even if modern state-of-the-art cameras were in place. Mass surveillance serves no useful purpose in North America, and would require public support and massive investment to accomplish. A country which has repeatedly rejected the use of a national identification card for more than a hundred years is unlikely to embrace mass surveillance. Yet police do use FR as a crime solving tool, usually when there are no suspects, but cameras have captured images. Then FR searches, highly accurate with portrait images but less so otherwise, may produce investigative leads, sometimes helpful in re-starting a stalled investigation. When considering the use of FR, it is important to remember that we do not have perfection today, but FR is a much more accurate method of identification than any known alternative.