Biometrics firms pitch privacy in age assurance ahead of US court battle

The U.S. is facing its first constitutional debate connected with age verification in 20 years: The Supreme Court will have to review Texas’ adult content age-verification law which requires that websites implement age verification to prevent minors from accessing inappropriate sexual material.

More than 20 states in the U.S. have passed similar legislation. The ruling on the Texas law, known as the Free Speech Coalition Incorporated v. Ken Paxton case, will affect the constitutionality of them all.

As legislators, free speech organizations and representatives of the adult industry share their arguments, biometrics, digital ID and age assurance companies such as IDVerse, Yoti and Incode are busy pitching their solutions.

“Particularly now, the demand for age verification is extending well beyond adult content and is particularly important for social media,” says Iain Corbin, executive director of the Age Verification Providers Association.

The organization has 30 members who supply the technology that either does traditional age verification or more novel age estimation techniques. Some of them presented their solutions at a webinar this week, from ID document OCR reading and facial recognition through selfies to more novel and unusual approaches such as facial estimation and age assurance with the movement of the hands.

“We think that facial biometrics is a crucial link that ties digital identity directly to a real-world identity,”  says Yoon Kim, Incode senior manager for sales strategy.

Many people, however, shy away from showing their real-world identity when accessing pornography, according to Terry Brenner, IDVerse’s head of legal, risk and compliance for Americas.

“As far as it comes to adult entertainment, porn sites, even social media, documents are not commonly used as part of the verification process,” says Brenner. “At IDVerse, we believe that is probably the most laser-sharp method of determining with accuracy the age of an individual, but we realize that there are different contexts in which it may or may not be appropriate.”

Accessing age-restricted content anonymously

While Incode showcased how liveness detection helps spot deepfake videos and images that can fool human eyes and IDVerse discussed the arrival of computer-generated, synthetic ID cards, Yoti and Privately explained how age assurance can be done anonymously – even without revealing a person’s age.

“[Facial estimation] gives consumers an accessible, fast, privacy-preserving way to prove their age, says Omari Rodney, chief product engineer at Yoti. “Everyone’s got a face. Everybody can use it.”

Yoti, which works on age assurance with companies such as Instagram, uses a neural network trained to recognize age by practicing on millions of faces. To build up this AI, the amount of data required is extensive and the firm has spent a long time acquiring data with user permission.

“It’s really, really hard to get hold of good data,” says Rodney “It’s really, really, really hard to get hold of good data in an ethical way.”

Yoti claims that their technology reduces the probability of somebody being estimated in the wrong age group to 0 percent. An age check takes one to two seconds on average while the images are not stored, shared with third parties, or checked against pre-existing databases.

The firm has performed over 600 million age checks across sectors like the adult industry, retail, telecoms, social networks, dating and gaming. The technique is also used by law enforcement to assess the ages of victims and perpetrators in CSAM and to support charities in the help of the removal of non-consensual content.

Yoti has recently released its facial age estimation white paper, showing that the True Positive Rate (TPR) for 13-17-year-olds correctly estimated as under 21 years is 99.3 percent, while the TRP for 6-12-year-olds correctly estimated as under 13 is 99.5 percent.

Swiss company Privately takes Yoti’s privacy concept to the next level. The facial image used to execute the age check never leaves the device, explains the company’s CEO Deepak Tewarey.

“The only signal going out of this age check is whether the camera saw a person who’s an adult or not and this changes the game,” says Tewarey. “If you think about how we can scale this technology, how it can be deployed across millions and millions of age checks, cost has been a big concern of the industry. If we were able to use the processing power in the device of the user to roll out age estimation quite broadly, this presents a great opportunity.”

Privately has created a multimodal solution relying on face or voice biometrics or a combination of both. Its voice-based age estimation technology, deployed by the adult industry in France to detect whether someone was below 15 years of age. Its anonymous age estimation technology was also embraced in retail.

Giving a hand to more unusual age assessment

The last age estimation technology presented at the webinar required some thinking out of the box. Jean-Michel Polit, chief business officer at Needemand introduced a creative new way of age assurance which is done by analyzing the movement of hands.

“It’s not magic, it’s physics or, really, it’s physiology,” he says.

Medical research has linked the movements of hands with age and the company has developed AI models with machine learning that leverage the age-specific features or certain hand movements. A user only needs to do a few hand movements and their age is verified.

“The first benefit is obviously that there’s no personal data in any way, shape or form that is required – not even metadata,” he says.

The second benefit is 99 percent accuracy, Polit adds. The software can detect attempts to fool the age assurance system with recordings of other people’s hands, hand mannequins and gloves.

Other event participants introduced alternative ways of confirming a person’s age. VerifyMy presented its email address age estimation method which detects age through user behavior, while Envoc explained how the U.S. state of Louisiana’s mobile driver licenses (mDL) can be used to access adult content anonymously just by typing a challenge code inside the mDL app.

More age verification solutions could be on the way as the U.S. Supreme Court deliberates on the Free Speech Coalition v. Paxton case. The decision is expected next year while interested sides can file amicus briefs until the 22nd of November, 2024.

Texas age verification law mostly unconstitutional: US district court judge

Article Topics

age estimation  |  age verification  |  AVPA  |  biometrics  |  data privacy  |  Envoc  |  face biometrics  |  IDVerse  |  Incode  |  Needemand  |  Privately  |  VerifyMy  |  Yoti