D-ID: how to fool facial recognition algorithms

Facial recognition artificial intelligence

*This writer contacted D-ID and had a chance to correspond directly with CEO Gil Perry for his insights on their breakout technology.

Facial recognition and artificial intelligence technologies actively transform digital photos and videos into biometric data that are used to identify and track individuals, raising serious privacy concerns in recent years.

To name just a few ongoing applications of this technology in the world today, retailers in the US apply facial recognition to monitor the behavior of shoppers, facial identification is used to create targeted marketing campaigns on social media, and some governments even apply facial recognition to track protestors and rank their citizens’ behavior.

Manage your supply chain from home with Sourcengine

To protect against potential data abuse, the Tel-Aviv technology startup, D-ID, recently announced their facial recognition protection software earlier this month at TechCrunch Disrupt SF 2018. Their software protects photos and videos from facial recognition by resynthesizing images to fool AI algorithms.

The catch? The images appear virtually indistinguishable to the human eye.

A Look inside the D-ID Team

D-ID founders (left to right) Eliran Kutam Gil Perry, and Sella Blondheim. Photo credit: PRNewsFoto.

D-ID was founded by CEO Gil Perry, CTO Elira Kuta, and COO Sella Blondheim in 2017.

The name D-ID is short for “de-identification,” supporting the company’s declared goal to defend identities and restore user privacy as consumer technology continues to grow more invasive.

Their inspiration to explore facial recognition privacy originated from their experiences serving in the Israeli Defense Force. During their service they traveled through exotic regions of South America, collecting daily photos unrelated to their operation, but were disappointed to find that security regulations prohibited them from sharing with friends and family.

Such privacy concerns on sharing sensitive biometric data led the D-ID team to pursue a solution to freely share photographs while remaining hidden from traditional facial recognition identifiers.

Current and Future Applications of D-ID

At the TechCrunch presentation, Perry explained that D-ID’s anti-facial recognition software was initially intended as consumer-focused, privacy protection for the public, but early in development, the company’s focus shifted toward larger organizations that have a more immediate need for the technology. Since most companies already store massive amounts of photos and videos, they have plenty of biometric data to protect.

The Burn-In asked Perry if D-ID had any plans to still make their software available to the general public for their personal use, to which he responded, “Yes, however, it is too early. Consumers can’t fully protect themselves now. First D-ID will be made available to the big social networks, phone manufacturers, workplaces, and governments. After we accomplish that, and if the demand keeps on increasing, we will provide a solution to consumers.”

D-ID has already started protecting the private data of their first customer, the image and video management website Cloudinary, which stores the photos of over 350,000 companies. This comes in addition to signed agreements they hold with clients in the finance and automotive industries who each have their own use for D-ID software. For example, as the technology in autonomous vehicles evolves, the cameras they use will capture the faces of pedestrians, and D-ID will be applied to keep their identities secure to prevent privacy violations.

In fact, a large portion of D-ID’s anticipated clients are companies who want to ensure that they uphold legal compliance with data protection laws to avoid fines and lawsuits. To meet that end, D-ID’s software is designed to abide by data protection and privacy regulations from around the world, such as the European GDPR, the US HIPAA, and Illinois’ BIPA guidelines.

Starting Ahead of the Curve

Previously, most attempts to avoid facial recognition were about wearing masks or specially applied makeup to fool cameras. Until now, nothing close to D-ID has been announced.

When The Burn-In asked about potential competitors, Perry notes, that there are only “…wearables such as the glasses that AVG developed which are supposed to blind cameras. Then, of course, there’s encryption, and there are methods such as pixelization and blurring.”

Perry went on to explain that there are other companies researching ways to prevent facial detection, but their progress is only partial compared to D-ID’s software performance.

“Our solution fools most classifiers and doesn’t require you to wear anything or hide. Also, unlike encryption, a photo that is D-ID protected is visually similar for humans, and cannot be decrypted or reverse engineered,” Perry states. “It is the first and only practical solution which was built for that exact purpose—to prevent face recognition seamlessly.”

The Forecast for D-ID

The spread of D-ID’s innovative technology will have important implications for the operations of countless organizations across the globe. With everyone from international businesses to world governments all storing image and video data, facial identification software is becoming more relied upon than ever before.

Perry speaking at TechCrunch Disrupt 2018 in San Fransisco. Photo Credit: TechCrunch

When we asked whether D-ID expects any backlash over how their technology could potentially disrupt the processes of major organizations, Perry stated, “It’s a good question, we have experienced this, but I prefer not to elaborate on this regard. We are not going to stop halfway until we make sure D-ID is available to all, and we have restored privacy. We also experience very good reactions from such companies, as they are the most aware about the sensitivity of the data they store and do not want this data to be breached and used by others.”

It’s evident that D-ID is optimistic about biometric data privacy, but what do you think the evolution of facial recognition software will look like in the future? Let us know in the comments below.

*You can find the D-ID onstage presentation from TechCrunch Disrupt 2018 here.