Discover how AI-powered tools are helping financial institutions detect fraud in real time and stay ahead of emerging threats like synthetic identities.
As financial fraud becomes more sophisticated, organizations grapple with new challenges in detecting and preventing scams. The rise of AI-powered tools has given fraudsters new ways to exploit digital vulnerabilities, making real-time detection crucial for financial institutions.
RTInsights sat down with Xin Ren, Head of Risk and AI for North America and APAC at Feedzai, to talk about how financial organizations are leveraging advanced solutions like Railgun to stay one step ahead (if not more) of these threats.
Interview has been lightly edited for clarity.
RTInsights: What are the emerging threats in financial crime, such as synthetic identities and money mule networks?
Xin Ren: Threat actors leverage the same technologies that financial institutions use, like machine learning and AI, but for fraud. Today, we see a shift where fraudsters are using generative AI (GenAI). GenAI is driving a major threat—the rise of synthetic identities. Fraudsters can create fake identities by combining real, stolen, or fabricated information, often sourced from data breaches and publicly available information online. This process is faster and cheaper than the traditional method of recruiting someone to open accounts manually.
With synthetic identities, the fraudsters don’t have to expose themselves, and they can create multiple fake identities quickly. . It’s incredibly difficult to detect because they’re stitching together all sorts of data—whether real, stolen, or fake—to essentially create a new person.. And with GenAI, they can do this at scale. They’re able to generate fake but convincing text, images, and even videos, which makes it extremely difficult for banks to spot inconsistencies and detect fraud, especially when traditional verification methods rely on things like name and address checks.
What makes this especially dangerous is that the best lies include some truth. You can’t just reject someone if their data isn’t a perfect match, especially in lending where you have to make some educated guesses. Even with sophisticated verification processes, fraudsters can use GenAI tools to generate voices or videos that match their fake identities, making it even harder to catch them.
RTInsights: What do organizations need? How are they using predictive AI and real-time data to detect these emerging threats?
Xin Ren: One of the biggest challenges for banks today is that they focus on monitoring money leaving accounts but don’t pay as much attention to money coming into accounts. This is a missed opportunity because fraud can be detected earlier by monitoring both sides of the transaction.
There are several things organizations need to do, and I’ll break it down into four key aspects.
First, it’s crucial to stop separating anti-money laundering (AML) efforts from fraud detection. Traditionally, these two areas operate as distinct functions, often with different reporting structures, but they shouldn’t. Fraud and money laundering are interconnected, and information needs to flow freely between these functions. Monitoring both where the money is coming from and where it’s going is essential to catching fraud before it becomes too difficult to trace.
Second, while most organizations focus heavily on KYC (Know Your Customer) during onboarding, many synthetic identities slip through the cracks. Fraudsters are becoming technically and operationally more sophisticated, making it difficult to detect fake identities until transactions occur. That’s why continuous monitoring, even after the onboarding process, is critical. It’s not enough to just monitor outgoing funds; you also have to look at incoming transactions..
Third, organizations need a more comprehensive view of their data. Traditional fraud detection focuses on where the money is going, but with the rise of synthetic identities, it’s essential to track money movement in both directions—both inbound and outbound payments—to understand the full picture.
Finally, it’s essential to look beyond just transactional data. You need to consider a broader set of information—biometrics, network data, login patterns, shared IPs or emails—should also be included. Fraudsters don’t just open one account at one institution; they often create hundreds of accounts, lying dormant until they’re ready to move money. By linking data across devices, networks, and biometric patterns, organizations can better detect fraud in real time and prevent significant losses before they happen.
RTInsights: How does Railgun help?
Xin Ren: Railgun is Feedzai’s patented technology. To explain it simply, think of traditional financial institutions—they typically rely on batch processing. For example, they might generate features—specific data points—every hour or run micro-batches every five minutes, which works well in many cases. But when you’re dealing with real-time data, especially with digital devices and fraudsters who move quickly, batch processing isn’t enough. If a fraudster spots an opportunity, they’ll move money out almost immediately—often within minutes. They don’t leave the money sitting around, especially in cases of ransomware or scams.
To catch this kind of activity, you need to process data in real time. That’s where Railgun comes in. Railgun enables real-time profile calculations, which is crucial for fraud detection. What makes it so powerful is its ability to work with unlimited windows of data. You could pull in five minutes of data, or even five years’ worth if needed. Of course, you don’t usually need five years, but the point is that it’s available if necessary. This real-time aspect is what sets Railgun apart.
Another key feature of Railgun is its ability to efficiently handle cross-channel data. One of the challenges in the financial industry is that much of the infrastructure was built decades ago, making it difficult to analyze events across multiple channels, such as digital activity and money movement, in real-time. While this older infrastructure is reliable, it wasn’t designed for today’s fast-paced digital environment.
Railgun addresses this by allowing financial institutions to create real-time profiles, holding various types of data in memory—whether it’s digital, device-related, or transactional information. This is cutting-edge for fraud detection. Companies like Meta and Google have already developed advanced systems for real-time data processing, but many financial institutions haven’t yet adopted this level of capability. Railgun helps close that gap by bringing similar real-time processing to the financial industry.
Railgun also helps integrate inbound and outbound information, providing a holistic view that allows real-time detection across multiple channels and signals, including digital devices and email. Ultimately, it helps you monitor both money coming in and going out, giving you a comprehensive view of money movement, which is critical for detecting fraud.
RTInsights: Can you share some examples of how Railgun has been put into practice in different organizations?
Xin Ren: Yes, absolutely. One example involves a large lender in the Asia-Pacific region. They’ve been using Railgun to aggregate data across various signals, such as digital activity and login data. By pulling all this data together in real time, they’ve been able to spot fraud and scams much faster and more accurately.
Another case involves a bank in the UK. They’re currently focused on addressing the new regulation around scam liability, and Railgun is helping them manage both inbound and outbound transactions. This broader, more holistic view is essential for detecting fraud more accurately, especially when it comes to scams and money mules, which are often connected.
What’s also exciting is how Railgun improves operational efficiency for these banks. With real time access to risk indicators and signals, their fraud teams can work more effectively. For example, previously, an analyst might not know whether a case involved account takeover, a scam, or a money mule situation. With Railgun, they can quickly pinpoint the issue, allowing them to ask the right questions. If it’s an account takeover, they might ask, “Are you really who you say you are?” But if it’s a scam or money mule case, they’ll need to dig into identity details to spot fake or mismatched information. This kind of tailored approach dramatically improves how cases are handled.
Another powerful feature of Railgun is its ability to uncover criminal networks. Once it integrates all the data—like payments, login activity, and digital signals—analysts can use tools like link analysis or graph analysis to uncover deeper connections. For example, analysts might notice patterns that point to multi-generational fraud rings. . What’s important here is not just seeing these patterns, but being able to act on them by creating alerts or models based on these discoveries to catch similar fraud cases in the future. That’s a huge benefit with major downstream impacts of using Railgun for new fraud detection.
Ultimately, it’s about getting that complete, real-time view across all channels. When you can do that, you detect fraud faster and help your teams ask the right questions, preventing scams and account takeovers before they cause serious harm—both financial and emotional.
