Facial Recognition Database Facing Potential Legal Action For Using Photos, Many of Children, Without Permission
The massive MegaFace dataset may have violated the Illinois Biometric Information Privacy Act, a 2008 law that protects residents from using facial scans without their permission.
- By Haley Samsel
- Oct 14, 2019
A facial recognition database holding more than 4 million photos of nearly 700,000 people is undergoing new scrutiny for its use of photos from Flickr without the express permission of users.
A New York Times report explores the relationship between the progress of surveillance technology and the availability of huge databases of facial photos on the web, including MegaFace.
MegaFace was developed by computer science professors at the University of Washington and consisted of downloaded versions of photos from the Yahoo Flickr Creative Commons 100 Million Dataset. The project was part of an effort to make it easier for smaller companies and researchers to further their development of facial recognition technology, among other goals.
The Yahoo database did not distribute users’ photos directly. Rather, links to the photos were shared so that if a user deleted the posts or made them private, researchers would no longer have access to them.
But MegaFace made the photo sets downloadable, making it easier for companies to download the data and use it for research purposes. The Flickr dataset was ideal because it had many photos of children, which facial recognition systems typically have a difficult time identifying accurately.
The University of Washington went on to host the “MegaFace Challenge” in 2015 and 2016, asking companies working on facial recognition to use the data to test the accuracy of their systems. More than 100 organizations and companies participated, the Times reported, including Google, SenseTime and NtechLab. All companies were asked to agree to use it only for “noncommercial research and educational purposes,” and some businesses said they deleted the dataset after the challenge.
Now, many of the people who posted photos of their children to the site say they were unaware that their children’s faces had been used to develop facial recognition technology. The data set was not anonymized, meaning that the Times was able to find people who had posted the photos through the links provided by Yahoo.
“The reason I went to Flickr originally was that you could set the license to be noncommercial,” Nick Alt, an entrepreneur in Los Angeles, told the Times after finding out that photos he had taken of children were in the database. “Absolutely would I not have let my photos be used for machine-learning projects. I feel like such a schmuck for posting that picture. But I did it 13 years ago, before privacy was a thing.”
Most people included in the database were not legally required to grant permission to use their photos because they were licensed under Creative Commons. But residents of Illinois are protected under the Biometric Information Privacy Act, a 2008 law that imposes fines for using someone’s fingerprints or face scans without consent.
The use of Illinois Flickr users’ photos could lead to legal implications if residents decide to pursue lawsuits. Photos themselves are not covered by the law, but scans of the photos should be, according to Faye Jones, a law professor at the University of Illinois.
“Using that in an algorithmic contest when you haven’t notified people is a violation of the law,” Jones said, adding that people who had their faceprints used without permission have the right to sue and earn $1,000 per use. That fine could go up to $5,000 if the use was “reckless.”
The combined liability could add up to more than a billion dollars, the Times reported.
“The law’s been on the books in Illinois since 2008 but was basically ignored for a decade,” Jeffrey Widman, an attorney in Chicago, told the Times. “I guarantee you that in 2014 or 2015, this potential liability wasn’t on anyone’s radar. But the technology has now caught up with the law.”
About the Author
Haley Samsel is an Associate Content Editor for the Infrastructure Solutions Group at 1105 Media.