Meta turns to a technology it once avoided to combat 'relentless' scammers (2024)

Meta is rolling out facial recognition on Facebook and Instagram to combat scammers who impersonate celebrities to promote products or investment schemes.

"Celebrity endorsement scams" cost Australians millions of dollars and feature well-known public figures including David Koch, Dick Smith, Mike Baird, Andrew Forrest, Sam Kerr and Dr Karl.

Meta turns to a technology it once avoided to combat 'relentless' scammers (1)

Meta has struggled to stop the spread of these scams on its platforms, despite deploying an automated monitoring system that uses machine learning to review millions of ads per day.

In recent weeks, it has tested an extra layer of monitoring that uses facial recognition technology.

It's now rolling out the technology to 50,000 celebrities worldwide.

The automated system compares images of public figures in suspected scam ads against the individual's Facebook and Instagram profile pictures, Meta's director of global threat disruption David Agranovich said.

"If we determine that there's a match and that the ad is a scam, we'll block it.

"Our goal is to make the platform more difficult for scammers to use."

Similar facial recognition technology will also be used to help users gain access to their account if they forget their password, lose their device, or when a scammer has tricked them into turning over a password.

How will the new features work?

The 50,000 public figures involved in the scam-monitoring trial will be sent in-app notifications letting them know they've been enrolled in an experiment and instructions for opting out.

"The types of public figures that we're including … have been historically impacted by this type of [scam] activity," Mr Agranovich said.

Public figures who aren't on Meta's platforms, but whose likeness is being used for endorsement scams, will not be part of the trial.

"They would have to have a profile in order to take advantage of this additional layer of protection," Mr Agranovich said.

Loading...

The facial recognition feature for regaining access to accounts will be rolled out more broadly to Facebook and Instagram users.

Unlike the celebrities selected by Meta for the spam-ad trial, users will opt in to the service when they try to regain access.

To do this they will upload a "video selfie" that will be compared, using facial recognition, against profile pictures on the account they're trying to access.

According to Meta, the process of recording a video selfie and having it verified against profile pictures will take one minute.

What happens to the data and user privacy?

Meta has a controversial history with facial recognition.

Three years ago, in November 2021, it shut down the "face recognition" photo-tagging feature on Facebook and deleted the facial templates of more than a billion people as part of a company-wide move to limit the use of facial recognition in its products.

The social media network agreed to pay a $US650 million ($970 million) settlement in 2020 after a lawsuit claimed the tagging tool violated Illinois privacy legislation. In July 2024, it agreed to pay $US1.4 billion ($2.1 billion) for using facial recognition technology without users' permission.

Mr Agranovich said Meta's latest rollout of facial recognition features had undergone a "robust" privacy review involving "regulators, experts, policymakers and other key stakeholders".

The anti-scammer facial recognition system will immediately delete any face data generated during the process of comparing the ad with the public figures' profile pictures, he said.

"We do not use the face data generated from ads or potentially violating accounts for any other purpose other than this one-time comparison.

"We immediately delete any face data generated from ads or potentially violating accounts, regardless of whether or not our systems find a match."

Loading...

For facial recognition used for account recovery, the video selfies will be "encrypted and securely stored" and not visible on the user's profile.

Meta will immediately delete any biometric data gathered from the facial recognition process, Mr Agranovich said.

"We don't use these video selfies for anything else other than verifying the identity and keeping our community safe."

Meta turns to a technology it once avoided to combat 'relentless' scammers (2)

But there may still be tricky questions to answer around privacy law for opt-out celebrity scam-ad monitoring.

Mark Andrejevic, a digital communications researcher from Monash University, said the default use of facial recognition for scam-ad monitoring raised alarm bells.

"The creation of biometric templates without permission has been found to contravene the Privacy Act," Professor Andrejevic said.

"It seems the creation of biometric templates of Australian celebrities and public figures without their permission would do the same."

Is the facial recognition accurate?

Facial recognition can have baked-in racial and gender bias, and be less accurate for women with darker skin in particular.

Mr Agranovich declined to share figures on the accuracy of Meta's facial recognition and the effectiveness of the broader automated scam-monitoring system.

"What we anticipate is this will enable us to catch violating content faster and, in doing so, we should be able to cut down on the amount of scam activity that we're seeing," he said.

"[The automated monitoring] process is done in real-time and is faster and much more accurate than manual human reviews."

Meta doesn't report how many scam ads it blocks and deletes per day, which makes it impossible for outsiders to judge the success of the facial recognition system.

Projessor Andrejevic said it wasn't clear why Meta needed to use facial recognition to combat scams.

"How do you habituate a society to the ever-more-pervasive use of facial recognition technologies? Saying, 'Oh, this will help us reduce scams' is one way."

He added that many scam ads on Facebook and Instagram expressly name celebrities so Meta should not need facial recognition to identify them as endorsement scams.

"It does raise questions as to whether facial recognition adds to the effectiveness of this system.

"Is this a necessary use of facial recognition technology or is this a bells and whistles marketing campaign?"

A spokesperson for the National Anti-Scam Centre, run by the Australian Competition and Consumer Commission, welcomed the use of automated systems to identify potential scams.

"This technology will facilitate faster identification of scams, supporting the National Anti-Scam Centre to quickly refer suspected scam URLs to take down services for disruptive action and alert the public to emerging scam types."

What's the scale of the celebrity endorsement scam problem?

We don't have firm figures, but it appears to be a growing problem.

Australians reported 18.5 per cent more scams of all types 2023 than in 2022, with reported losses of $2.74 billion, according to the National Anti-Scam Centre.

In October, Meta said it took down 8,000 celebrity endorsement scam ads from Facebook and Instagram.

Kathy Sundstrom, national outreach manager at the cyber support service IDCARE, said Australians were reporting more scams on social media platforms.

Celebrity endorsement scam reported losses were over $14 million in the first nine months of 2023.

"We had 437 reported cases. Of those, 247 were on Meta platforms, with Facebook being the lion's share.

"It's really important the platforms do something when it comes to celebrity endorsement scams. We hope this is the beginning of many different measures."

Mr Agranovich said scammers were "relentless".

"Since the pandemic, there's been a global rise in scams and attempted fraud," he said.

"[Scammers] constantly adapt their tactics to evade our enforcement and they aren't limited by borders, by countries or by online service or platform.

"While we have automated detection systems that are running against ads that are being created and do remove a large volume of violating ads … scam networks are very highly motivated to just keep throwing things at the wall in the hope that things get through.

"And invariably some of them do.

"Even if [facial recognition monitoring] is successful, scammers will probably migrate to other tactics."

Science in your inbox

Get all the latest science stories from across the ABC.

Your information is being handled in accordance with the

ABC Privacy Collection Statement.

Meta turns to a technology it once avoided to combat 'relentless' scammers (2024)

References

Top Articles
Latest Posts
Recommended Articles
Article information

Author: Errol Quitzon

Last Updated:

Views: 5742

Rating: 4.9 / 5 (59 voted)

Reviews: 90% of readers found this page helpful

Author information

Name: Errol Quitzon

Birthday: 1993-04-02

Address: 70604 Haley Lane, Port Weldonside, TN 99233-0942

Phone: +9665282866296

Job: Product Retail Agent

Hobby: Computer programming, Horseback riding, Hooping, Dance, Ice skating, Backpacking, Rafting

Introduction: My name is Errol Quitzon, I am a fair, cute, fancy, clean, attractive, sparkling, kind person who loves writing and wants to share my knowledge and understanding with you.