This Startup Is Using Facial Recognition to Fight Human Trafficking
3 hours ago Share
During a trip to Eastern Europe when she was 16, Emily Kennedy, founder and chief executive of startup Marinus Analytics, was passing through a small town with just a few stoplights. At one of those lights, a group of children slammed up against her car, clamoring to wash the windows.
It made her uneasy.
“After we passed through, a friend of ours who was from the area told us that those kids were trafficked by the Russian mob to beg on the street and get tips and that if they didn’t make enough money by the end of the day to meet their daily quota, they would be punished,” Kennedy said, adding, “I was confronted by the concept that this exploitation was happening to kids my age and much younger.”
Kennedy sought to learn more about human trafficking and became interested in how technology facilitates and but can also help fight the crime.
Per the International Labour Organization’s figures, an estimated 40.3 million people were forced into a modern form of slavery—including forced labor and sexual exploitation—in 2016, meaning there are 5.4 victims for every 1,000 people in the world. One in four victims was a child.
Trafficking is a $150 billion industry worldwide, and about a $1 billion industry in the U.S.—with missing, runaway and foster children the most vulnerable population. According to the National Center for Missing and Exploited Children, or NCMEC, an estimated one out of six endangered runaways in 2016 was likely a victim of sex trafficking.
Hannah Rivard, an officer with the Fort Worth Police Department, said stats are hard to come by but that she’s seen studies that say a single massage parlor in Houston can make $100 million a year.
The rise of sites like Craigslist in 1995 and Backpage.com in 2004 made it easy for traffickers to advertise online. Kennedy said traffickers want control, so when they worked with victims on the street, they usually hung around nearby, were in more potential danger themselves and couldn’t remain as anonymous. The internet, however, makes it much harder to tie any illegal activity back to the perpetrator.
It also makes it easier for johns.
“On the internet, you can buy anything,” Rivard said. “When it went to girls, it just made it that much easier for the johns to distance themselves. … They don’t have to know where to go on the streets … and get their face out there and drive around and risk being found by the police. … Do we have good stats on the numbers from [before the internet] versus now? No, of course we don’t, but it certainly provided a huge, huge market.”
Technology has also yielded better tools for law enforcement. That includes Marinus’ Traffic Jam, which Kennedy started developing when she was a student at Carnegie Mellon University. It’s a suite of AI tools, including facial recognition, that came out in 2013 and helps identify victims. It now has a database of over 210 million ads. The Traffic Jam tool helps law enforcement track escort ads so they can find trafficking victims faster. Marinus Analytics
“Every day, tens and even hundreds of thousands of escort ads are posted online,” Kennedy said. “We scrape the top escort sites and put them into Traffic Jam to make them searchable. The goal is to take all of the massive amount of data on the internet that’s relevant and turn it into actionable intelligence.”
That data includes phone numbers, locations and now photos after Marinus added a feature called FaceSearch in 2017.
FaceSearch uses artificial intelligence, machine learning, computer vision, predictive modeling and geospatial analysis. This allows detectives to start with a photo of a missing child to determine whether a potential victim has been advertised online.
The Fort Worth Police Department has been using Traffic Jam for about a year and a half, along with other software including Spotlight, which Rivard said provides more graphs and charts, including spider charts, that illustrate how phone numbers from ads are connected.
Detectives previously relied on phone numbers to track victims via Google searches, but traffickers use burner phones and get new numbers frequently, making such searches more challenging.
“This is an extremely slow and tough manual process that would not for most cases show you all of the ads,” Kennedy said. “One, because of human error, but also because traffickers delete old ads, so they are not on Google and can’t be found.”
Rivard agreed, but noted Traffic Jam crawls sites every 20 minutes, meaning it captures almost all the ads and puts them into a database.
“It literally makes cases we couldn’t make otherwise,” she said.
Law enforcement can also use intel from Traffic Jam to conduct rescues much faster—per Kennedy’s figures, it can reduce investigation time by 50 percent, which can work out to days or even weeks. Sites where ads appear are also a moving target.
In April, federal authorities seized Backpage.com. That meant Marinus had to find new data sources—Kennedy would not disclose the new sites but said some are not based in the U.S., which makes prosecutors’ jobs even harder. ( A recent story in the New York Post, however, said these sites include Canadian website Bedpage.)
While Rivard said Backpage certainly wasn’t perfect, it was at least friendly to law enforcement and responded to subpoenas with official records of ads.
“When it shut down … it’s not that the trafficking ended—it got dispersed,” Rivard said.
And, in a few months, it will likely settle on one main site again, she added.
But Marinus and Spotlight aren’t alone in their efforts. Just last week, Facebook hosted its third child-safety hackathon in which nearly 100 engineers and data scientists reportedly from companies like Uber, Twitter, Google, Microsoft and Pinterest worked on projects to help combat child sex trafficking. All code and prototypes were donated to partners NCMEC, Thorn, the Internet Watch Foundation, Stop the Traffik, A21 and Polaris.
Kennedy said Marinus works with local, state and federal law enforcement in the U.S. and Canada as well as some nonprofits, including the NCMEC and the DeliverFund. It has also worked with prosecutors and attorneys general and said she’d love to work with financial groups.
“We have all this data, and we’re able to now work with agencies to see overlap between online [ads] and the wire transfers of the proceeds of human trafficking … helping tie sex trafficking to other cases,” Kennedy said. “It’s often not isolated and is really connected with other crimes.” Share