FaceSearch, one of the tools included in the Traffic Jam AI suite, can take a photograph of a missing person and use facial recognition technology to potentially match it to photos included in ads on sites used by sex traffickers. (Photo: Courtesy of Emily Kennedy, Marinus Analytics)

Editor’s Note: Welcome to my weekly column, Virtual Case Notes, in which I interview industry experts for their take on the latest cybersecurity situation. Each week I will take a look at a new case from the evolving realm of digital crime and digital forensics. For previous editions, please type “Virtual Case Notes” into the search bar at the top of the site.

A teenage girl has gone missing. Police begin an investigation. Investigators fear she is being trafficked for sex. They post her photo online, and on flyers throughout the city.

There’s a chance someone might recognize her, and if officers have the time and resources, they might manually scroll through the thousands of escort ads posted online every day in the hopes of eyeballing a match. But with the vast size of the internet, and with the possibility that the victim has been moved far away from where she first went missing, finding that one face can be like trying to find a needle in a haystack.

Emily Kennedy, who was an undergraduate student at Carnegie Mellon University half a decade ago, had heard stories like these, and since encountering trafficked children during a trip abroad as a teenager, had made it a personal mission to help law enforcement combat the sale of human beings. Beginning in 2011 at Carnegie Mellon’s Robotics Institute, Kennedy pursued this mission using a technology that could scan for images, make connections and pull together data using much less time and effort than a human would need: artificial intelligence.

“There was a huge amount of data, hundreds of thousands of new ads selling sex every single day,” Kennedy said in an interview with Forensic Magazine. “It was really just trying to make the connection between the huge amount of data, and then the stories that we had heard about, patterns that existed and ways that individual detectives had found victims—but thinking about a way that that could be scaled up and made more automatic.”

In 2014, Kennedy became the CEO of her own startup Marinus Analytics, which offers a suite of AI tools called Traffic Jam that law enforcement agencies across the United States, and, recently, Europe and Asia, have been using to find and rescue trafficking victims, and prosecute their traffickers. Among the suite of tools are two image search functions—FaceSearch and SimSearch—that replace the endless scrolling and eyeballing of detectives with an automatic process that takes seconds.

With FaceSearch, detectives upload a picture of a suspected trafficking victim, and the tool uses facial recognition technology to scan ads from sites like Craigslist and Backpage to see if the victim is being advertised as an escort. FaceSearch has a success rate of about 88 percent in finding positive matches to victims, according to Kennedy. She described one example from San Francisco, in which a 16-year-old girl had gone missing.

“[A detective] took her picture, which was pictured in a public news article, uploaded it to FaceSearch, and immediately found a top match, went to those ads, [and] determined that she was in a city in the San Francisco area,” Kennedy explained. “It looked like her and she had just started advertising around the time that she went missing. And he was able to work with the local agencies to go out and conduct a successful rescue and arrest two of her pimps for human trafficking, and this is in the span of a week.”

Emily Kennedy is the CEO and founder of Marinus Analytics. In February, Kennedy was awarded the Toyota Mother of Invention honor and a $50,000 grant for her work using AI to help law enforcement track down victims and perpetrators of human trafficking. (Photo: Courtesy of Emily Kennedy)

Kennedy said she has heard many similar success stories from investigators using Traffic Jam. And beyond discovering ads of people who have gone missing, Traffic Jam can draw quick connections between known trafficking ads to potentially bust a trafficking ring, as well as see where perpetrators and their victims may be moving.

“What SimSearch does is it looks in the foreground of the image, which is typically a person, what’s in the background of the image, which is typically a hotel room or a bathroom, and it returns the most similar images based on either the foreground or the background or both,” Kennedy explained. “What it’s essentially able to do is find multiple victims pictured in the same location, in the same bathroom or on the same bedspread for instance.”

Another tool, Trail, can compare one ad to the 200 million ads in Traffic Jam’s database, looking for similar wording, contact information and pictures. Kennedy said these connected “trails” have consisted of anywhere between 10 and over 2000 ads, and these connections are made in just seconds as opposed to the hours or days it might take a detective to make them through their own investigation.

“You just need one ad, one perp’s ID off of a page, and you can start making your whole case,” said Officer Hannah Rivard, of the Forth Worth Police Department in Texas, to Forensic Magazine. Rivard said the department began using Traffic Jam shortly after she joined its human trafficking unit about 18 months ago. “We’ve had several cases where we have found [ads] via SimSearch, via a Facebook photo that we had. I can’t tell you how many times we’ve printed out ads and confronted traffickers with their ads right in the interview, in real time.”

Kennedy said she has received feedback that Traffic Jam replaces the work of a full-time analyst, and added that the tools overall reduce investigative analysis time by about 50 percent. She also said she receives lots of constructive feedback from agencies, which can help improve Traffic Jam—for example, she said that sometimes the simplest searches for names, phone numbers or locations in the database can be as valuable to investigators as the hi-tech recognition capabilities of FaceSearch. This kind of feedback can guide the company on where to focus to best help law enforcement.

Kennedy recently received the Toyota Mother of Invention award, and a $50,000 grant to help with her work. Kennedy said she plans to use the grant to help expand Traffic Jam’s user base, to spread the word about the AI suite to more agencies and hire someone to help with user growth and support.

Rivard said she is looking forward to seeing how AI can continue to improve investigations, and hopes to see it extend beyond just human trafficking cases.

“There’s no reason to still be in the Stone Age when doing investigations,” Rivard said.