ADOT's facial recognition technology leads to arrest of identity theft suspectPosted: Updated:
Just one day after a Casa Grande woman walked into an MVD office for a driver's license, investigators officially launched an identity theft investigation against her -- all thanks to facial recognition software.
Sharon Forrest, 51, has been charged with 10 felony counts related to forgery and identity theft.
Court records show that when Forrest had her driver's license photo taken on Feb. 2, the Arizona Department of Transportation's NeoFace software determined her face closely resembled two other ID photos in the database and flagged the images.
Detectives with FBI training reviewed the photos the following day, and determined all three images were of Forrest. Investigators said she had been using two stolen identities to get Social Security benefits, Department of Economic Security financial assistance, settlement payments and veterans benefits.
Court documents indicate Forrest first obtained an Arizona ID card under a false name in 2008. She obtained another Arizona ID card in her cousin's name in 2012.
She was arrested March 1.
"This tool helps us find those identity thieves much quicker and be able to pursue those cases much faster," said ADOT spokesperson Ryan Harding.
ADOT has been using NeoFace, a software program developed by NEC Corporation of America, since 2015, Harding said.
Like other image recognition software, NeoFace uses a relatively new form of artificial intelligence called "deep learning."
"Deep learning is the idea that you can teach a computer to learn a pattern by starting with small, lower-level structures like edges, and then from those small structures, learn more complex patterns," said Nadav Ben-Haim of Orpix, another image recognition software company.
"You start off with edges and colors and build to eyes and noses," he said. "The really remarkable thing is that in the last few years, that kind of technology has basically allowed computers to become more accurate and better at this recognition of faces than humans."
ADOT still uses humans, however, to review each potential case of fraud. Photos undergo "three levels of review" by detectives who have received FBI training, the agency said in a statement.
“Humans are involved at every level of facial recognition technology,” said Michael Lockhart, chief of the Office of Inspector General. "We make sure we’ve got solid evidence before advancing one of our cases. This is not a science fiction scenario where the computer determines potential fraud all by itself.”
Copyright 2017 KPHO/KTVK (KPHO Broadcasting Corporation). All rights reserved.