This software differentiates authentic images from deepfakes

03 May This software differentiates authentic images from deepfakes

According to a 2022 study, human headshots created by artificial intelligence have now crossed the uncanny valley, meaning that we are now more likely to think a fake face is real than an authentic one. Synthetic media—photos and videos created by tech—are now so advanced and prevalent that deepfakes are no longer just the work of expert digital crafters. Thispersondoesnotexist.com, a “random face generator” website, shows that phenomenon, conjuring up images of real-looking people who don’t actually exist.

This worrying democratization of synthetic media is the motivation behind Truepic Lens, a software development kit (SDK) that can fully integrate with apps that rely on images for their operations, allowing them to verify media in real time, and provide assurances to customers where necessary. Truepic also envisions a deeper cultural importance: to restore trust in a world rife with disinformation. When anything could be fake, what can be trusted? “We’re now looking at this future of: How do we operate where synthetic media is available to anyone?” says Jeffrey McGregor, Truepic’s CEO.

Truepic Lens, the winner of the software category of Fast Company’s 2022 World Changing Ideas awards, is a kit that businesses across industries can install, “and empower them with that trusted layer,” McGregor says. In real time, Truepic Lens validates all the metadata of an image, including time, date, location, and device—then cryptographically signs it, and embeds the validation in the file. It provides that certification to businesses on the back end, and also to customers if desired. It’s compliant with C2PA, a new set of standards for content verification authenticity.

While Truepic is not yet releasing names of its upcoming business partners, it provides some application examples. It could help Airbnb to verify host images, proving they’re not pulled from a Zillow ad; allow Uber to ensure it’s advertising genuine drivers; and reassure Bumble users that they are not being catfished. “The entire digital economy is at risk if trust evaporates,” McGregor says. Insurance companies will also be able to save time—and carbon emissions—digitally corroborating people’s claims, instead of making in-person trips.

McGregor compares the system to a traditional notary, and emphasizes the importance that Truepic is a neutral third-party. He’s also reluctant to sell it directly to Apple or Android. They don’t want to just be available to the newest iPhone, “or this particular flavor of Android on this device,” he says. “You want it to be universal, across the entire ecosystem.”

Pre-SDK, Truepic Vision was authenticating images using controlled capture technology, for which it won a World Changing Ideas award in 2019. In that original iteration, they worked mainly with financial services companies, but also allowed for the virtual validation of PPP loans during the pandemic, and the verification of war crime footage from conflict zones like Syria and Yemen, among a handful of journalists.

Now, Lens creates a way to scale all those applications. “With Lens, we can take that same technology, and put it in the BBC app,” says Mounir Ibrahim, Truepic’s vice president of strategic initiatives. “Anyone in the world, if they’re taking a picture and sending it to a bank, or taking a picture and sending it to The Hague, will be trusted.”

Business aside, the company’s other ambitious aim is to recreate trust in a society teeming with polarization, much of which is due to rampant disinformation. “Truepic is not the arbiter of truth,” McGregor says. It can verify what’s truthful and what’s not, without weighing into contentious political debates. “This general distrust that has permeated society,” Ibrahim adds, “that’s where we hope to be very, very impactful long term.”

Source: Fast Company