News > i3

Facing Up to Tech: A Look into Facial Recognition and Biometrics


Whenever federal policy makers begin scrutinizing a fast-evolving technology, especially when they threaten legislation, it sets the stage for a faceoff between regulation of the unknown versus innovation into uncharted opportunity.

That duel is taking shape in the realm of biometric technology, exemplified by facial recognition technology (FRT), which has encountered a slew of congressional initiatives and agency proposals, such as a Federal Trade Commission’s plan to require multi-factor authentication such as  facial recognition technology FRT or fingerprints in certain digital financial transactions.

Research also indicates that many consumers favor FRT’s frictionless security processes over passwords or other complicated digital access tools. A Veridium survey in February concluded that 70 percent of consumers would like to see expanded use of biometric authentication. The respondents cited speed, security and not having to remember passwords, especially on multiple devices, as the primary reasons for their enthusiasm.

All this comes as the government’s technology experts confirm there has been a 20-fold improvement in facial recognition software quality between 2014 and 2018. The National Institute of Standards and Technology’s (NIST) evaluated 127 software algorithms from 39 developers and identified the massive jump in the ability to find a matching photograph and mate its biometric algorithms. NIST’s test found that “just 0.2 percent of searches failed this year, compared with a four percent failure rate in 2014 and five percent in 2010.”

A bipartisan group of senators established the Senate Artificial Intelligence (AI) Caucus to examine facial recognition as part of its effort to complement the “American AI Initiative” launched by the White House early this year.

This flurry of activity came on the heels of a landmark agreement among Amazon, Microsoft and Google to support FRT regulations. While skeptics questioned whether their endorsement is an effort to head off congressional restrictions, the industry approval of a coordinated tech policy underscores the growing impact of facial recognition and biometric data. Amazon’s support of rules emerged soon after the company faced criticism for its “Rekognition” FRT technology that is used in applications ranging from military to marketing.

Michael Punke, vice president, Global Public Policy at Amazon Web Services, explained his company’s stance on its blog.

“Over the past several months, we’ve talked to customers, researchers, academics, policymakers and others to understand how to best balance the benefits of facial recognition with the potential risks,” Punke said. “It’s critical that any legislation protect civil rights while also allowing for continued innovation and practical application of the technology.”

His comments added to the growing dialogue about the role of FRT. For example, Rep. Emanuel Cleaver, II (D-MO), chairman of the House Subcommittee on National Security, International Development and Monetary Policy has focused on the social implications of FRT. “Facial recognition is a powerful tool that is permeating American life, and yet, the propensity of the technology to misidentify individuals, particularly in regard to variances in skin type and gender, is welldocumented,” Cleaver said. “The potential for illegal discrimination and/or unfair practices resulting from such bias continues to concern lawmakers.”

Beyond the Brouhaha: Growing Implementation

While policymakers ponder how to handle FRT and its allied biometric tools, countless communications, retail, academic and marketing organizations have embraced the capabilities of these technologies. Its financial impact is expected to grow from $3.8 billion in 2017 to $9.8 billion by 2023.

Whether it’s smartphone authentication (such as Apple’s Face ID or Android’s Smart Lock systems) or classroom facial monitoring (to assure that only authorized students are in the classroom as well as to take attendance), FRT has established itself as a viable and often vital technique.

Going farther, research companies are using “emotion recognition” to fine-tune advertising campaigns and other creative ventures by observing how facial responses such as smiles or grimaces indicate viewers’ attitudes toward specific images.

In hospitals, FRT is being used for quicker, more accurate patient care. Online retailer Alibaba’s “Smile-to-Pay” software has been used for fast-food purchases for nearly two years. Snapchat lets users establish privacy settings by using FRT to determine which friends should be blocked by an emoji. Facebook’s system has helped its members tag friends when they appear in photos for nearly eight years.

HyreCar, a car-sharing marketplace for ridesharing now operating in three dozen states, introduced a mobile application in March that can expedite new driver onboarding and verification, using technology from software developers Mitek. HyreCar CEO Joe Furnari says that the new seamless mobile interface can expedite the booking and verification process by 30 percent and lower ID document forgery up to 20 percent — all of which will “enable scale and reduce overhead costs toward our future vision of personal transportation.”

Fundamentally, FRT blends computer vision and data processing to measure and match dozens of unique facial characteristics; the most basic systems capture about 80 nodal points to create a faceprint (a numerical code) that analyzes features such as the length of a jawline, the shape of a nose and cheekbones and the distance between eye sockets. Subsequently, this image, reduced to data, is put into a facial database and can then be compared to other images. For example, in security authorizations, the current face scan is matched to the image in the database to confirm that it is really you.

In most configurations, these FRT recognitions can be integrated with existing security systems, such as employee or student ID. Experts acknowledge that processing and storing visual IDs can be burdensome, but they observe it is another cost of doing business.

The recent Biometric Consumer Sentiment Survey, commissioned by Veridium, a developer of authentication solutions, identified the growing appeal of such systems. The introduction of fingerprint sensors into mobile devices in 2011 was the catalyst for the consumerization of biometric authentication, says Veridium CEO James Stickland. The study found — perhaps reflecting familiarity with current systems — that fingerprint ID was preferred by 63 percent of respondents, compared to facial recognition (14 percent), voice recognition (two percent), or traditional passwords and PINs (eight percent).

Veridium also found that consumers are comfortable using biometrics to unlock their devices (80 percent), and to use applications such as finance (35 percent), payments (31 percent), company networks (12 percent), travel (11 percent) and health care (10 percent). Among the demographic preferences uncovered in Veridium’s study:

Millennials (under age 35 years) value speed (46 percent), Generation X (ages 35-55) cite not having to remember passwords (44 percent) and Baby Boomers (55+ years) favor security (30 percent) more than anything.

  • Millennials most frequently use biometrics to access financial applications such as banking apps or ATMs (46 percent) and payments (45 percent). Generation X’s number one application of biometric authentication is for travel (41 percent) and Baby Boomers most use the technology for health care applications (28 percent).

 
  • Millennials are more likely than other generations to say they haven’t been the victim of a data breach (46 percent). Generation X is more likely than any other generation to have been the victim of a data breach (47 percent).

Dramatic Tech Improvements

The ongoing policy tension about expanding facial recognition and other biometric technologies corroborates the impact and improvements throughout the category. For example, the recent NIST analysis found that all of the top-performing algorithms use machine-learning software architectures called “convolutional neural networks.” Patrick Grother, a NIST computer scientist and one of the report’s authors, observed that the results underscored the advancements in the technology.

“The test shows a wholesale uptake by the industry of convolutional neural networks, which didn’t exist five years ago,” Grother said. “About 25 developers have algorithms that outperform the most accurate one we reported in 2014.” But he added, “There remains a very wide spread of capability across the industry.”

Separately at CES 2019, dozens of companies exhibited devices and systems that use FRT for biometric tracking of products ranging from home and family security to health monitoring. Sensory, a Santa Clara-based firm, demonstrated applications of its technologies (face authentication, voice and speech recognition) that have been licensed into more than one billion consumer electronics devices.

iLumintel, based in Chengdu, China, and San Jose, CA, showed its 3D behavior recognition technology “iLu-IDTracer,” a 3D facial image capture and behavior recognition “tracking people flow” system, which goes beyond static mug shots. The system can follow individuals in any complicated environment, such as monitoring employees moving around a store or stockroom or keeping track of a customer’s entry and exit times and browsing behavior/pattern while in the store.

RealNetworks known for its music and other technologies, has been developing SAFR (pronounced “safer”) — Secure, Accurate Facial Recognition — and began offering it for free to kindergarten through 12th grade schools last year. It is an off-shoot of the company’s RealTimes app that lets customers build photo slide shows. The technology maps 1,600 data points on every face and took more than two years to develop. In the process, RealNetworks also spent three years to identify eight million faces for SAFR and more than eight billion data points, with an eye toward increased school safety.

Elsewhere, products such as Apple’s iPhone X “Face ID” exemplify the rapid improvements in 3D facial recognition. The iPhone’s infrared camera collects data from more than 30,000 invisible dots on a face to confirm a face; then its software can recognize a face in various lighting conditions and from different angles as well as with changes such as wearing sunglasses or changing your hairstyle.

Facial recognition technologies inevitably trigger discussions that go beyond their impressive capabilities. Given the global attention to such security, safety technologies, especially in politically sensitive areas such as China, the current focus by U.S. policymakers is part of the eye-popping process.

Illustrations by Stephanie Dalton

Gary Arlen

Tagged

Related