Cognitive computing and geospatial technology help law enforcement track, locate, and rescue human trafficking victims
The Intelligence Community has created powerful tools for monitoring potential bad actors in the War on Terror—software that can help identify a terrorist, discern when multiple online presences are in fact the same person, visualize an individual’s movement through time and space, and reveal terrorist networks. Increasingly, these same technologies are being applied against human trafficking, particularly child sex trafficking in the United States, to rescue children and put their traffickers behind bars.
Approximately 325,000 U.S. children are at risk for sex trafficking and 75 percent of these victims are advertised online, according to Thorn, an organization dedicated to driving technology innovation to fight the sexual exploitation of children. With vulnerable children not only targeted and recruited online but also sold there, the ability to ingest and analyze vast amounts of unstructured data is paramount to bring them to safety.
Thorn was founded in 2009 by actors Ashton Kutcher and Demi Moore, who began by learning as much as they could about the issue, according to Thorn CEO Julie Cordua.
“During that exploratory process, it was very apparent that technology was increasingly a part of these crimes against children, but there was no concentrated effort to make technology a part of the response,” Cordua said.
As technology further enables human traffickers, more and more organizations are recognizing technology must also be a part of the solution.
“The use of forums, chats, advertisements, job postings, hidden services, etc., continues to enable a growing industry of modern slavery,” said Wade Shen, a program manager with the Defense Advanced Research Projects Agency (DARPA).
DARPA, Thorn, and others have created and continue to improve platforms that mine data and make connections among nefarious material on the web. As government contractors and Silicon Valley tech companies retrain their tools on urgent and complex social issues such as human trafficking, they experience victories as well as encounter new challenges.
Cracking the Code
Many law enforcement agencies lack the funding to dedicate officers to human trafficking cases, let alone to purchase trafficking-specific software. Those working in the field described the pursuit of intense cases with a lack of resources.
“If you start calling random police departments asking, ‘Who works human trafficking?’ you’re probably going to get 50 percent of agencies that say, ‘Nobody,’” said 2nd Lt. James Bacon, who oversees the Child Exploitation Squad for the Fairfax County Police Department (FCPD) in Virginia.
Yet according to Kate Reilly, a special agent with Homeland Security Investigations (HSI) New Orleans whose primary mission is to recover juveniles involved in sex trafficking, such crimes are prevalent in all major U.S. cities. Even with three or four officers at a time dedicated to human trafficking in New Orleans, resources are stretched.
“We’re constantly being pulled in a million directions,” Reilly said.
Both FCPD and HSI New Orleans employ software such as Thorn’s Spotlight—a platform leveraging big data analytics and visualization, machine learning, and natural language processing—to automate the monitoring of ads on websites such as backpage.com. Thorn, in collaboration with technical partner Digital Reasoning, provides Spotlight for free to federal, state, and local law enforcement agencies around the country.
Digital Reasoning, a boutique software company in Nashville, Tenn., got its start in 2000 through In-Q-Tel funding and spent more than a decade primarily supporting the defense and intelligence communities in the War on Terror. Today, in addition to its work with the financial sector, Digital Reasoning has invested considerable resources in its Spotlight cognitive computing platform, which reimagines the company’s terror fighting tactics to be applied against human trafficking.
Thorn and Digital Reasoning joined forces in 2014, after Thorn learned in its interviews with survivors that many trafficked children were forced to write their own online ads. Using both supervised and unsupervised learning, Digital Reasoning trained Spotlight to determine which ads were likely written by children and elevate the most high-risk ads to help law enforcement focus their investigations.
According to Bacon, software such as Spotlight helps police conduct link analysis to detect networks and track movement from one location or jurisdiction to another.
“It can be very exhaustive to try to manually search through these things, so to have an automated process or piece of software that can go through and do analysis and give a data dump and say, ‘Based on this search criteria, you need to be looking in this direction’ is huge,” Bacon said.
Spotlight’s unstructured text analysis goes beyond identifying key words to mine human communication for certain concepts and behaviors.
This approach allows Spotlight to detect code words and keep up with trafficking language as it evolves. While there might be a lag between the introduction of a new word and its identification by law enforcement, Spotlight’s unsupervised learning capabilities allow it to understand semantic and syntactic context to identify words used similarly to others of interest to law enforcement.
“If I’m going to bribe you, I’m never going to send you an email with the word ‘bribe’ in it,” said Eric Hansen, vice president of federal programs for Digital Reasoning who spent 20 years in Army intelligence. “Same in the terror [or trafficking] communities. Words are used in a way that is non-obvious.”
Spotlight is now in the hands of more than 3,000 U.S. law enforcement officers at 780 agencies across all 50 states, according to Cordua. As of September 7, the platform has aided in 8,305 trafficking investigations, assisted in identifying 2,025 child sex trafficking victims and more than 4,624 adult victims, and helped bring more than 2,249 traffickers to justice. Thorn will soon introduce Spotlight in Canada and plans to expand its international capabilities in coming years.
Spotlight’s natural language processing not only reveals the most high-risk advertisements but also the links among them.
“There may be a series of ads that mention teddy bears,” Hansen said. “In one set of ads, in one locale, it may say ‘my teddy.’ In another series of ads in a different locale it may refer to ‘bear.’ A lot of times we can understand it’s the same person writing the ad—one time in Arizona, another in Kansas. Now you can see geospatial movement.”
Perhaps more obvious than language, visualizing the use of the same phone numbers in different locations helps reveal the full scope of a trafficking situation.
When OGSystems’ Viper Labs sought a real-world use case to test the ability of its new DeLorean app to ingest, process, analyze, and display open-source data, it learned law enforcement manually searched for concerning ads and common phone numbers on sites such as backpage.com. The DeLorean team created a spider to crawl the website, ingest ads, process and structure their data, and then use location information to geotag them.
“The visual display would automatically draw lines that connected an ad in Washington, D.C., to an ad in Chicago using the same phone number,” said Aaron Snow, a consultant with OGSystems. “We took a very manual process for investigators, automated it, and provided a geospatial element to it.”
This in many ways mimics how federal intelligence agencies have automated analysis—enabling law enforcement agents to read for confirmation rather than comprehension.
Similarly, Spotlight’s geospatial capabilities can provide powerful historical context to a single ad, according to Cordua.
“Over time as well as across space you see geographically where [the person posting has] been and also the network that person is connected to,” she said.
This automated process might, for example, allow an investigator trying to find a child advertised in Los Angeles to learn the victim is a runaway from San Francisco that has been trafficked up and down the California coast. Or an officer who has only been on the job five years to link newly posted ads to a trafficker who was jailed 10 years ago and recently released—using methodologies characteristic of activity-based intelligence.
“That’s what’s cracking some of these cases open,” Hansen said. “Creating links that never would have been created using the manual process.”
The analysis teams at the National Center for Missing and Exploited Children still conduct much of their research manually, according to Angela Aufmuth, program manager for the center’s Special Analysis Unit. But the organization is working to streamline processes and is increasingly realizing the importance of location data, she added.
“[Human trafficking] is such a transient crime,” Aufmuth said. “We know there’s so much moving of victims—being able to track that and have those visualizations is important and certainly helpful for law enforcement.”
Special Agent Reilly said she uses Thorn in two primary ways. One is to track the history of a specific ad. For example, if she recovers a 15-year-old girl from an ad with a phone number, the software can help her learn when and where else the girl has posted. The second is mobile alerts.
“If I’m home or out in the field, getting those alerts is really helpful,” she said. “If I’m trying to track a specific girl, I can say ‘She’s posting, what can we do?,’ mobilize, and go forward with a plan operationally.”
DARPA takes analysis of human trafficking activity on the web a step further with its Memex program. Memex connects dots among sex trafficking information available via commercial search engines such as Google; the “Deep Web”—internet accessible to the public but not indexed by common search engines; and the “Dark Web”—the corners of the internet that obfuscate users and where illegal marketplaces thrive.
“By enabling searches across a wide range of websites, [Memex] uncovers a wealth of information that might otherwise be difficult or time-intensive for investigators to discover,” Shen said. “Possible trafficking rings can be identified and cross-referenced with existing law enforcement databases, which helps police officers and public prosecutors map connections between human trafficking and other illegal activity.”
Law enforcement agencies and district attorneys’ offices around the country are using Memex. For example, the Manhattan district attorney’s office employs the tool for all of its human trafficking investigations, and in Texas Memex has been used in more than 200 cases leading to more than 100 arrests.
Communication between developers and end users is key when leveraging technology for humanitarian causes.
Cordua said she is proud of Thorn’s development process and that the idea for the platform came from understanding the victim mindset and how they move through time and space.
“But nothing else matters unless the person who is trying to go find them can use the tools,” she said. “You have to understand what the user needs.”
In the beginning, Cordua said, many law enforcement officers Thorn met with didn’t think the organization’s goal was possible.
“We got a lot of people who didn’t buy in—who said, ‘No, you can’t. It’s not going to be helpful.’ But we kept working. We showed we could,” she said, adding that an intuitive user interface was essential to enabling ease of adoption for law enforcement.
Although investigators aren’t currently using OGSystems’ DeLorean spider, the company is exploring opportunities to tailor the tool more specifically to the needs of law enforcement agencies for future operational use.
And increasingly, organizations are standing up to help bridge the cultural gap between technologists and activists. ATHack!, co-founded by Ehb Teng, is a San Francisco-based social impact venture aiming to do just that.
“We realized there were two big divides as we delved into the issue [of human trafficking],” Teng said. “On the nonprofit side we see a digital divide—technologically they aren’t as up to date. And on the tech side we are finding a lack of understanding of actual, critical social issues that need to be solved.”
In April 2016, ATHack! co-hosted their ATHackathon with Microsoft Reactor to bring the two sides together. Nonprofits were invited to talk about specific projects they needed help with so technologists could develop mid- to long-term solutions.
Melissa Jane Kronfeld, founding co-chair of the Nexus Working Group on Human Trafficking, spends much of her day connecting tech entrepreneurs’ capabilities with nonprofit organization and law enforcement needs. Nexus is an invitation-only, global organization of about 3,000 millennial philanthropists and innovators.
“Technology and data are the future of fighting human trafficking because so much of it has moved online,” Kronfeld said. “Whichever way you look at it, data and technology are how we will end slavery.”
While analytic software has powerful potential to enhance the efforts of human trafficking investigators, there are still gaps technology has yet to address.
The internet has democratized abuse, with many vulnerable children being recruited online—often by other children already being exploited. In addition to rescuing children faster, part of Thorn’s mission is to keep them out of harm’s way to begin with by making online platforms safer and stopping underage users’ risky behavior.
The ability to intercept online recruiting is a need the OGSystems team noticed as well.
“[Online predators are] pervasive, and technologies that try to match these patterns and do the geotagging and really do the research on social networks at a rapid rate would have a higher probability of achieving some success in this area,” Snow said.
Beyond analytic software, law enforcement is particularly interested in whether geospatial technology could help locate exploited children via their smartphones. The first question Reilly asks when looking for a child is, “Do they have a phone?”
Location-based technology in smartphones makes it easy to determine a person’s whereabouts should their phone connect to WiFi and therefore an IP address tied to a specific location.
“But if a kid is logging onto the internet from the phone’s [data plan] and all we have is a T-Mobile IP address, we can’t triangulate that data down to a location,” Reilly said.
According to both Reilly and Bacon, cellphone carriers claim they do not yet have a way to triangulate that information.
“[Trafficked children] are able to conceal their activities much easier than it is for us to track them,” Bacon said.
Both officers indicated this is perhaps their greatest technological challenge— one they would welcome outside solutions for.
The anti-trafficking community is just at the beginning of exploring how technology can help end modern slavery, according to Hansen.
“The genie is out of the bottle,” he said. “We can now read human communications and do it at a massive, massive scale … [Spotlight] is cutting edge coming out of the IC and is really the tip of the iceberg if you think about things like Twitter, Facebook, and the amount of human language being ingested every day.”
Jason Beck, director of communications for Digital Reasoning, said although the company’s work with the IC and the financial sector is rewarding, the staff is energized when it hears success stories from the anti-trafficking community.
Beck said, “To hear from a detective that, because of the analytics we were able to provide and the targeting we were able to help with, a 12-year-old girl was assisted and her [victimizer] is now in jail—those are the kinds of things that are remarkable.”