Federated Learning Promises to Change the Game for Money Laundering
Federated Learning Promises to Change the Game for Money Laundering (solutionsreview.com)
There’s a new sheriff in town to stop money launderers, and it’s a subset of Machine Learning known as Federated Learning (FL). This technology has the ability to analyze data without actually moving the data, so for the first time ever financial institutions can collaborate without violating privacy regulations and the evolving landscape of expectations.
Why is this a big deal? Let me explain: criminals place, layer, and spend illicit proceeds through numerous financial intermediaries to avoid detection – this is the essence of money laundering. These bad actors are usually not as smart as portrayed in movies, but because the system they exploit is so vast that catching a money launderer is the equivalent of finding the proverbial needle in the haystack. This vastness only increases as dark web and crypto assets further expand the domains of illicit trade.
In my years in national security and law enforcement, I collaborated every chance I could. But for commercial institutions such as banks and fintechs, identifying criminal money launderers and human traffickers and maintaining customer privacy appear directly at odds. For example, financial institutions could discover more money laundering if they could share data with each other and across jurisdictions, but privacy regulations included in Europe’s GDPR and virtually every U.S. state law forbids this practice.
Federated Learning, however, breaks this historic tradeoff. Financial institutions can now, for the first time in history, collaborate and simultaneously protect privacy.
Risks, Technology & Regulatory Conundrum
FL is now available to the Financial Crimes Compliance market. Over the past several years, the cost to organizations to comply with regulations and identify money launderers has skyrocketed due to two factors. First, the regulatory burden expands each year and banks are required to do more in a complex market that includes e-commerce, fintechs, crypto, dark web, etc. Second, regulated financial institutions continue to use the same tools and technologies they’ve been using for years. Most of these tools were created decades ago and were never meant to cope with today’s world. Criminals use the latest innovative technologies to evade discovery while the banks fail to adopt innovations, or move much more slowly.
This conspiracy of a growing regulatory regime and the innovation gap between criminals and bankers creates many problems, one of which professionals in cybersecurity are only too familiar with: false positive alerts. These old anti-money laundering (AML) systems throw off between 97-98 percent false positive alerts, meaning that financial crime employees spend most of their days chasing down could-be money launderers that aren’t. It is no wonder that turnover in this industry remains high. Plus, criminals understand what these AML systems are capable of, so they are careful not to do things that will trigger them and know it is rare for them to be caught.
Enter Federated Learning
To dramatically decrease false positives, find more risk, and spend less time and money, financial institutions must collaborate. FL enables collaboration without moving data – thus protecting privacy.
Think of cyber and financial crimes as patterns in data. Criminals are human, and humans behave in largely predictable ways. Through FL, people can exploit this behavioral science insight. And they can share this insight with external organizations – whether they are other offices in different jurisdictions or different companies. That’s where FL platforms come into play – they enable organizations to share information without violating privacy.
The Power of Moving Analytics to Data
The next time a cyber-crime discussion gets hijacked by data-sharing concerns, shift the discussion to sharing the analytics. By moving algorithms to data through FL, models can continue to improve and evolve, and these insights can be shared with others without sharing any data.
Technology adoption in crime fighting usually lags that of the criminals. It’s a matter of incentives – criminals look to exploit holes that enable them to make money, and the “good guys” reactively patch those holes to stop the criminals. Fortunately, the era of FL is slingshotting the good guys ahead. Think about it: criminals can adopt AI and move to crypto and other places with alacrity, but FL allows financial institutions to more easily collaborate and stop them. By dramatically reducing false positives and increasing discovery of malfeasance, FL allows financial institutions to finally put a dent in the dirty money-laundering industry.
