How Amazon Rekognition helps in the fight against some of the worst types of crime
A detective sits at a computer with a victim’s picture taped to the screen as he scrolls through pages of online ads looking for a visual match. That’s how police investigators described for researchers at Carnegie Mellon University how they search for missing children.
By the close of 2018, the FBI’s National Crime Information Center Missing Persons File for children under the age of 21 had 38,561 open records. Of the more than 23,500 runaways reported to the National Center for Missing & Exploited Children in 2018, one in seven were likely victims of child sex trafficking.
Today, with online ads, social media, and smartphones, sex trafficking can happen virtually anywhere. But it remains hidden from the general public, and law enforcement struggles to keep up with the pace of the crime.
"It has definitely changed the landscape," former detective Rich Lebel said of the increased use of internet advertising to attract clientele. In fact, the rise of “invisible” human trafficking through online media has changed the way officers must investigate such cases. Lebel now uses Amazon Rekognition in his work to identify and locate victims, then collaborate with law enforcement to conduct recoveries.
Machine learning for good
While certain corners of the internet have become a haven for traffickers, the cloud-based technologies that Amazon and other companies are developing are part of the toolkit for those fighting trafficking. Tech-powered social impact organizations such as Marinus Analytics are building software solutions with services including Amazon Rekognition to analyze data and expedite human trafficking investigations. Such solutions ultimately help reunite missing children with their families and solve cold cases, which Lebel notes, “couldn’t be done” without such technology.
Emily Kennedy co-founded Marinus Analytics in 2014 after her research at Carnegie Mellon on human trafficking illuminated a pain point for detectives: mountains of data and no efficient way to analyze it.
Marinus integrated Amazon Rekognition within its suite of tools in 2017, enabling investigators to find victims much more quickly by using facial recognition technology to search through millions of records in seconds, even in the case of grainy or poorly lit images.
"We saw success pretty much immediately," Kennedy said, pointing to a case where a two-year-old social media photo, matched to online ads, helped surface the victim of a violent trafficker and put the criminal behind bars.
In his current role at a trafficking awareness nonprofit, former detective Lebel uses a tool that Marinus Analytics developed on top of Amazon Rekognition (called Traffic Jam) to interpret data for law enforcement. “Traffic Jam uses Amazon Rekognition to not only help locate victims, but also to detect and establish patterns among a sea of online ads that are difficult, time-consuming, and at times distressing to search and review,” he explains.
From his experience, the facial recognition technology helps find victims faster, which enables law enforcement agencies to recover a victim within days or weeks rather than after months or years have elapsed.
"Amazon Rekognition helps us by identifying other sex ads that we didn't know about, which in turn leads us to additional telephone numbers that we didn't have before," he said.
“Not using these tools would be harmful for the victims we’re trying so hard to help.”