Artificial intelligence, or AI, is quickly becoming a key tool in the fight against an evolving, global human trafficking business that is worth an estimated US$150 billion a year, especially as the pandemic has highlighted the ability of perpetrators to adapt quickly and continue to remain hidden, say researchers and bankers working to disrupt financial crimes.

Advances in technology, such as AI graphing capabilities — using the technology to create graphs and relationships or connections between transaction data — help create context around transactions in relation to other connected or similar data, said Stuart Davis, EVP, Financial Crimes Risk Management & Group Chief AML Officer, Scotiabank.

“Without technology, we could not find that needle in a haystack,” he said.

While banks are starting to leverage AI and machine learning to connect customer information and transaction records to detect human trafficking activities, it’s still in the early stages, says Carrie Chai, Director, AI and Machine Learning, AML/ATF Risk Models.

Chai is part of the team that developed two AI-based, industry-leading models for Scotiabank: one to detect human trafficking related to sexual exploitation; another to detect child sexual exploitation.

These models use sophisticated machine learning techniques and leverage enriched transactional data − such as withdrawals made online or at automatic banking machines, credit card use, and electronic money transfers − as well as external data to create context and connectivity to detect potential human trafficking criminal rings and suspicious activities related to child sexual exploitation.  

“It’s usually a combination of red flags and suspicious transactional behaviours in the context of a network or criminal ring that give the whole story,” Chai said.

Human trafficking and child sexual exploitation activities are complex and difficult to build models for. A combination of AI and machine learning techniques are applied in these models in order to capture these suspicious activities.  Natural Language Processing (NLP), which understands text similar to how humans do, is used to group similar occupations and detect sensitive information from email addresses and transaction memos. Lastly, network and graphic analysis are used to detect different roles and structure with a human trafficking criminal organization such as traffickers, handlers, victims and buyers.

Academic research

For the past 18 months, Professor Uyen T. Nguyen and her students have been developing financial technology to assist banks in detecting suspicious transactions and financial patterns that lead to potential illegal activities. Scotiabank’s donation late last year of $980,000 over four years to the Lassonde School of Engineering at York University will help fund their research through the Scotiabank Lassonde Financial Crimes Research Initiative, as well as help boost diversity and equity in STEM (Science, Technology, Engineering and Math).

Nguyen’s path to financial crimes technology began when a personal interest in finance and investment collided with a network security project on peer-to-peer (P2P) networks she and her students were working on.

“The work on P2P networks began as a way to help dissidents in countries ruled by a dictatorship communicate with colleagues, friends and family. At the same time, I was thinking I may be helping criminals evade law enforcement, the same way dissidents and activists circumvent censorships. That’s how I got into researching technological loopholes criminals involved in human trafficking are exploiting,” she said.

The Lassonde researchers use financial technology, as well as an integration of different kinds of technology — machine learning, big data, social media, cloud computing and cyber security — to arrive at solutions, said Nguyen, a professor of computer science at Lassonde School of Engineering at York University in Toronto.

The partnership will help advance some of the models and techniques used to detect human-trafficking-related crimes, Davis said. At the same time, the investment will help train the next generation of researchers. “That will have a huge impact not just here in Canada but hopefully globally and specifically in our efforts on anti-human trafficking and online child exploitation,” he said.

Sharing data without sharing information

Getting data to plug into the models is key to advancing the technology and training new machine learning models, but with the increased use of data, regulators and financial institutions are paying more attention to how privacy is being protected. Scotiabank, for example, does a privacy impact assessment on the data it uses, and has added an ethical questionnaire to the process to ensure data is used responsibly and ethically.

Financial institutions have also been looking at how they can share information between banks without revealing personal identifiable information (PII).

“The criminals are breaking up their money laundering activities across multiple institutions to try to hide it. We are only seeing a part of any story, and until you connect the dots with external information, you may not even know there’s a crime going on,” Davis said.

If we can leverage cutting-edge encryption and data matching technology, sharing data without sharing PII could be a reality sometime in the near future, said Chai. She sees data sharing as a huge win in the fight against human trafficking and child exploitation because it would give banks better grounds on which to raise suspicious cases with regulators. This richer data would be fed into the Bank’s existing AI detection models and produce more robust suspicious transaction report (STRs), she added.

New activities require new models

Canada’s financial intelligence unit, Financial Transactions and Reports Analysis Centre of Canada (FINTRAC) requires reporting entities that submit STRs to provide an explanation for how they came to that conclusion. Currently, most machine learning models only produce classification results without explanations and sorting through large amounts of data to find the reasons behind the results is time-consuming, Nguyen said. “Even the designers of the algorithms cannot explain why the machine arrived at a particular conclusion,” she said, adding that research using explainable AI — a new technique that enables machines to provide the explanation along with a result — should make reporting easier.

One of the challenges Nguyen hopes to tackle is to quickly and effectively revise the indicators for identifying human trafficking activities and, in turn, develop AI models to reflect the new operational environment of human traffickers.

Amid the pandemic, human traffickers and criminals quickly moved their business from in-person to online in the face of prolonged lockdowns, Nguyen said. At the same time, the number of children and adults using social media has risen significantly, giving traffickers an opportunity to recruit new victims, she said. Nguyen also pointed to the increased use of cryptocurrency, which makes it more difficult to track suspicious activities because each transaction is not associated with a specific user but rather with a crypto address.

The role of public-private partnerships

AI models aren’t the only tools being used to disrupt human trafficking and child exploitation. In Canada, public-private partnerships (PPPs) such as Project Shadow, co-led by Scotiabank and the Canadian Centre for Child Protection (C3P), FINTRAC and law enforcement, primarily focus on raising awareness and developing common typologies and methods that can be collectively used to detect financial transactions related to modern-day slavery and exploitation.  The typologies, or red flags, are then shared publicly through an Operational Alert published by FINTRAC, which mobilizes all reporting entities, such as banks and other financial entities, covered by the regulator.

On February 22, Canada marked National Human Trafficking Awareness Day. February 22 has been designated by the government to bring awareness to the magnitude of modern-day slavery in Canada and abroad and to encourage Canadians to take steps to combat human trafficking.

“Raising awareness with the public is super important. I think we can all do our part,” Scotiabank’s Chai said.

“Financial institutions are only part of the solution. Human trafficking is a global problem but also one happening in our own backyard, and it requires the help of not just modelers or AML professionals to disrupt,” she said.