Meta is expanding its testing of facial recognition as a counter-fraud measure to combat fraudulent celebrity ads and more broadly, the Facebook owner announced on Monday.
Monika Bickert, Meta's vice president of content policy, said in a blog post that some of the tests include enhancements to existing fraud protection, such as automated scans (using machine learning classifiers) performed as part of the ad review system. I wrote that it was my purpose. It makes it harder for scammers to fly under the radar and trick Facebook and Instagram users into clicking on fake ads.
“Scammers often use images of public figures, such as content creators or celebrities, to try to trick people into ads that direct them to fraudulent websites asking them to share personal information or send money. “This scheme, called 'celebrity baiting', violates our policies and is harmful to those who use our products,” she wrote.
“Of course, celebrities appear in many legitimate ads. However, celebrity bait ads are often designed to look authentic, so it's not always easy to detect them. ”
The test is a backstop to check that existing metasystems mark ads as suspicious if they contain images of public figures at risk of so-called “celebrity baiting.” It seems to be using facial recognition.
“We will use facial recognition technology to compare faces in ads to celebrities' Facebook and Instagram profile photos,” Bickert wrote. “If we see a match and the ad is confirmed to be a scam, we'll block it.”
Meta claims that this feature is not used for any purpose other than to combat fraudulent advertising. “Regardless of whether the system finds a match or not, the facial data generated from the ad for this one-time comparison will be immediately deleted and not used for any other purpose,” she said.
The company said initial testing of the approach with “a small group of celebrities and public figures” (it didn't say who) showed “promising results” in improving the speed and effectiveness of detecting and cracking down on this type of fraud. He said that the results were obtained.
Mehta also told TechCrunch that he believes the use of facial recognition could be effective in detecting deepfake fraudulent ads, where generative AI is used to generate images of celebrities.
Social media giant uses its advertising platform to stop scammers from exploiting celebrity faces to scam unsuspecting users into dubious cryptocurrency investments and other scams It has been criticized for years for not being able to do so. So it's interesting timing that Meta is trying to capture as much user data as possible to train commercial AI models, while at the same time pushing facial recognition-based anti-fraud countermeasures to this problem. a broad industry-wide scramble to build generative AI tools).
Meta has announced that in the coming weeks, it will show in-app notifications to more celebrities who have fallen victim to Celebrity Bait, letting them know they are in the system.
“Celebrities who have enrolled in this protection can opt out at any time in their Account Center,” Bickert said.
Meta is also testing the use of facial recognition to spot celebrity fraudster accounts (e.g., accounts where fraudsters attempt to impersonate celebrities on the platform to expand their fraud opportunities). There is. This is also done by using AI to compare the profile picture of the suspicious account to the profile picture of the suspicious account. Facebook and Instagram profile pictures of celebrities.
“We hope to test this and other new approaches soon,” Bickert added.
Unlock your account with selfie video and AI
Additionally, Meta is testing the use of facial recognition applied to video selfies to help people unlock their Facebook/Instagram accounts more quickly after they have been hijacked by scammers. We announced what we are doing. (I was tricked into giving my password.)
This appears to be aimed at appealing to users by touting the obvious usefulness of facial recognition technology for identity verification. Meta suggests that it's a faster and easier way to regain access to your account than uploading an image of your government-issued ID (which is the usual method). Root to unlock).
“Video selfie verification expands the options people have to regain access to their accounts, takes just a minute to complete, and is the easiest way for people to verify their identity,” Bickert said. says. “While we know that hackers will continue to exploit account recovery tools, this verification method will ultimately be less exploitable by hackers than traditional document-based identity verification.”
The facial recognition-based video selfie identification method being tested by Meta requires users to upload a video selfie, which is then processed using facial recognition technology to identify the profile photos and videos of the account they are trying to access. Compare.
Mehta claims this method is similar to identity verification used to unlock a phone and access other apps, such as Apple's FaceID on iPhones. “As soon as someone uploads a selfie video, it is encrypted and stored securely,” Bickert added. “It will never be visible to your profile, your friends, or anyone else on Facebook or Instagram. Any facial data generated after this comparison will be immediately deleted, regardless of whether it's a match or not.”
Conditioning users to upload and save selfie videos for identity verification is one way Meta will expand its services in the digital ID space, if enough users opt-in to upload biometrics. Could be a method.
No testing in UK or EU – at this time
According to Meta, all these facial recognition tests are being performed all over the world. However, the company pointed out quite prominently that it is not currently being tested in the UK or the European Union, where comprehensive data protection regulations apply. (In the specific case of biometric authentication for identity verification, Block's data protection framework requires explicit consent from the parties for such use cases.)
Given this, Meta's test appears to be part of a broader PR strategy it has launched in Europe in recent months aimed at putting pressure on local politicians to weaken public privacy protections. This time, the cause for free data processing for AI is not a (selfish) notion of data diversity or a claim of lost economic growth, but a more direct goal of fighting fraudsters.
“We are working with UK regulators, policymakers and other experts on our testing,” Meta spokesperson Andrew Devoy told TechCrunch. “We will continue to seek feedback from experts and make adjustments as the feature evolves.”
However, while the use of facial recognition for narrow security purposes may be acceptable to some, and indeed may be tackled by meta under existing data protection rules, commercial AI Using people's data to train models is a completely different story.