The Current State of Fraud & Financial Crime

Speaker A: This is Leading Detection, a cedulo podcast where fraud fighters talk tech. Welcome back to the Lead in Detection podcast. And today we're joined by Colin Parsons. He is VP and head of Fraud Product Strategy at Nasdaq Verafin. Colin, thank you very much for joining us.

Speaker B: Hey Matt, thanks for having me. I appreciate it. Yeah, it's good.

Speaker A: So we'll be talking about the current state, uh, of kind of the fraud space. And we've got some data and a report to go through. You've got a great vantage point about what's going on. But before we get into the data, um, what would you say the mood is right now amongst fraud teams and fraud leaders?

Speaker B: Yeah, so right now it's, I think, a really interesting time in the fraud space. We, uh, you know, we spend a lot of time talking about, talking to our customers and working with financial institutions and fraud leaders. And there, there's a little bit of a, kind of a uncertainty, I'd say, with, uh, everything that's happening from an AI point of view. I think there's a lot of changes happening from the point of view of the way that fraud's occurring, the different types of attacks that we see, uh, in the space. And then also on the kind of detection and prevention side, there's just AI has kind of kicked off a change on both sides of the equation. So definitely some uncertainty, I think also some, uh, some optimism, you know, depending, depending on the day. But, uh, definitely, definitely some optimism in uh, in the tools that we have to, I guess, the newer tools that we have to fight fraud now. So, yeah, uh, it's, it's definitely an interesting time and happy to be in the space to work through this.

Speaker A: Yeah, I mean, we shouldn't disregard kind of feelings and instincts, but it's not as important as, uh, the data. Right. The instinct gu. Guides us. The data kind of proves or disproves what we're feeling. So, yeah, you just released a report on the current state of financial crime. So. Yeah. Do you want to run through some of the headlines from that report?

Speaker B: Yeah, for sure. So, yeah, we released our global Financial crime report and this has been something that we've released every. Every other year or this is the second version of it, I guess. And really it's. The goal is to help understand kind of the scope and scale of financial crime in around the world. And there the. The thing when I look at the report, there's some very large numbers which are sometimes hard to just wrap your head around. We saw that illicit financial activity that includes Fraud and money laundering has, has risen to 4.4 trillion um, annually which is up from it's is up 1.3 trillion in just two years. So at a high level, you know again those numbers are very large and like I said, hard to kind of wrap your head around. The increase alone over the last two years is kind of an indicator of the direction overall. We also see specifically on the fraud side of things, 579 billion in fraud losses. Fraud and scam losses. So again a really large number. And, but, but when I, when I look at the report the things that I kind of focus on are the changes over time. And one of those changes that we saw in the last couple of years was ah, specifically on the authorized fraud or scam side of things where um, that actually increased at a rate of 19%, 19.3% actually over the last two years which interestingly is double the rate of unauthorized fraud. So those kind of account takeover scenarios.

Speaker A: Mhm.

Speaker B: While still increasing. The authorized scam scenarios are what's increasing at a much faster rate. I think even when we just talk to our customers and talk to others across the industry. The, the shift while, while again those account takeover scenarios, people losing their credentials, that stuff is still a huge issue. But I heard somebody say that it's, it's not so much stolen credentials, it's stolen confidence. It's the, the scenarios where somebody is convinced a uh, victim to send money. That's really the challenge in the, in the industry right now.

Speaker A: Yeah, I mean it's with fraud prevention. It's not an or equation, it' equation. It just keeps accumulating. So looking at the fraud and the scams in particular authorized unauthorized from, from a high level, what would you think's triggering this kind of increase in the numbers?

Speaker B: Yeah, so in, so as part of this data, but as well as we actually recently had a group of our customers together in like a round table. So got a lot of feedback from those, the leaders across a whole bunch of different financial institutions. And in both cases we're seeing that uh, AI is driving a lot of the change here. So what our customers are seeing and what we see in the kind of the data here from the report is that it's a change from the kind of loan fraudster or small fraudsters to an uh, industrialized, uh, kind of machine driven view of fraud and scams where they're, they're taking advantage of AI to just be more efficient. Like you know, like every financial uh, institution, every organization is trying to do right now. They're doing the same things and it's driving, it's, it's making the, in particular the scam scenarios much more convincing, much more believable and much more frequent. So it's uh, you know, it's, it's changed it from that kind of human scale fraud to a machine scale fraud and really industrialized I think over the last number of years. And interestingly, when we talked about our customers and um, other financial institutions, they've really seen that, you know, I'd say probably a year ago the kind of we, we all saw the stories of like a deep fake and you know, a bank losing a single large transaction or, or something like that. But now it's a consistent, it's a regular thing that everybody is, is seeing almost on a daily or, or weekly basis and it's affecting you know, some of the smallest credit unions, smallest banks to all the way up to the tier ones. So uh, it's been an interesting shift, but yeah, it's taking the frauds and the scams that we've seen before and just really accelerating them.

Speaker A: Yeah, because we've seen the Brad Pitt romance scam, we've heard about the junior finance individual who joined a meeting with a CFO and others and wired, I think it was US$25 million. These are the headlines, right? And some people may think, well these are uh, edge cases. But it keeps coming up as a factor in AI. So is it sophistication? Is it volume, automation? Especially with these bank leaders? Well, how are they seeing it?

Speaker B: Yeah, it's, I think honestly it's a, it's a combination of all of those things. The challenge is, the challenge that I see kind of constantly come up over and over again is that it's not necessarily kind of the perimeter of the financial institution that's really being targeted. The, the kind of surface area has, has really broadened. So it's, it's all the way from a, you know, the point of compromise could be somebody trying to purchase something online or an interaction they have in on um, social media or through WhatsApp or something like that. And then the, the kind of, the ultimate outcome of the, that entire scenario is a financial loss like a, a transaction that gets sent to a mule account or a uh, some sor. Financial loss at the end of the day. But by the time it gets to that, that actual financial loss, that, that transaction point, so many things have already happened that the, you know, the, the really, the sad stories that we see all the time is these individuals are just locked up in the, the scams, they've been convinced over months or weeks that everything that they've seen is, is true. And that's come to a point of a, of a, of a uh, transaction or a financial loss. So I think that makes it harder for banks to kind of combat at the point of transaction. So then they need more insights than just what they can see in their own data.

Speaker A: Yeah, and this probably goes hand in hand with the unauthorized authorized fraud. So I mean if banks, customers were machines and scams wouldn't maybe wouldn't happen. But that's not the way the world is. Right. The human is the customer. So why do you think criminals have shifted focus away from the cyber side and more to the human scam side?

Speaker B: Yeah, yeah. I mean honestly, we've spent a lot of effort over the last, you know, 20 years securing the like, like I said, securing that perimeter and putting a lot of things in place. Again, it's always, it's an ongoing batt. There's, there's, there's always going to be changes or you know, uh, there's, there's always going to be credentials that are lost or stolen or data breaches or things like that, that, that will happen. So that's a constant battle. But I think we've gotten a lot better at that kind of unauthorized scenario.

Speaker A: Um, um.

Speaker B: And to be quite honest, the uh, fraudsters follow the path of least resistance. They will find the, the point in the, in the chain that's easiest for them to compromise. And right now that is, it's, it's the, it's the individual. Especially, you know, as it, as more and more it's hard to tell if something is, is real or fake, then that's, that's, I think going to be, going to be the point of compromise. That's, that's a challenge. But uh, but yeah, but it's, you know, all of those things are challenging. There's a lot of uh, like I said, I think uncertainty, um, in where the things are going. But I do feel like there's, there's still ways that we can, we can educate, we can provide different ways to identify these scen and, and, and prevent fraud and really protect individuals from losing money.

Speaker A: Yeah. And education. Right? Uh, how are you seeing the kind of the time to educate versus the time to change with fraudsters? Right. Cause that's changing so quickly. Yeah.

Speaker B: Yeah. I, so, uh, education. I heard a couple people say that friction is good if it's the right amount of friction at the right time. So putting friction in place is beneficial. It makes people feel like they're a part of the solution. They're, you know, they see that okay, my financial institution is helping protect. And then maybe as, as part of that friction there's education that comes along with it. Um, there's things that are happening up front so people are made aware of these scenarios and all of that makes the, the individual um, feel like they're a part of the solution and kind of carried along on the journey. When you get to those scenarios where it's ah, you know, you're, you have these things that you just need to click through to say yes, I meant to send this or yes I know who this is going to. Yes I'm aware those are the kind of scenarios where uh, it's, you know, it's like the fatigue of multifactor authentication or something like that where people just click through it and don't pay attention. So having, having the, the, the right friction as well as the like what I would kind of describe as like personalized education. So this scenario that you're currently in, this looks like a scam because of the value is unusual. You don't usually send this type of payment. Uh, the person that you're sending it to is you know, in a different country that you've never sent to before. This account looks like it was just opened. Those, those pieces of information become valuable from a detection and prevention point of view, but then also from like an education point of view and giving that information to uh, help, help uh, help inform the individual that they are potentially in a scam. Because again the, the scenarios that, that, that we hear from investigators is that they, they, they're, they're, they're wrapped up in these scams and it's so hard to kind of break them from that spell that they need as much data as possible to, to be able to uh, to inform the, the individuals rather than just, you know, I guess preach that they're potentially in a scam because they're not going to be open to that.

Speaker A: No, uh, and humans are irrational as that's the way we are. And sometimes that hey, if someone's had a newborn baby and they're shopping on a strange website for two in the morning when they wouldn't usually that might flag a fraud detection system. But if you have more data then you can make sense of that. So how are we working on getting data, uh, getting content, not just content, but getting context as well and doing that on kind of a uh, network level.

Speaker B: Yeah. So one of the approaches that Verifyn's been a big proponent of over the last 20 plus years is that network approach for any financial crime. We're all in this together. We are all. Any financial institution uh, is trying to uh, prevent the same fraudsters from attacking one to attacking another. So it's, and ah, it's a team sport and being able to have a view outside of kind of your own four walls as a financial institution allows, allows us to ideally be in a bit of a better place. Just fundamentally I think we spent a lot of time focusing on uh, who sent the money is what's normal for their activity. But really in this era of scams and authorized push payment fraud, where the money is going and understanding who the receiver of that payment is is, is a lot more valuable to actually preventing. So yeah, the, the, the value of the, the network is, is huge. When you're at an individual bank level, the, the bank kind of just sees the transaction. Uh, but the network sees that whole pattern of behavior and through like uh, through our analytics and, and our, our products we're able to then provide that insight back to an individual instit. Hey, this is the network view. This is what the network's seeing. This is the risk associated with this scenario based on the entire view across the network rather than just kind of your slice of the pie. So all of that view from one institution to another is, it's where the fraudsters kind of hide, they hide in that the lack of visibility from one institution to another. But being able to kind of um, uh, uncover those patterns and provide those insights back allows us to fight back a little bit and again identify the scenarios that are actually risky rather than just saying this doesn't look normal for you.

Speaker A: M. Yeah. And with fraud there's such a small window, right. And we're getting all this data, uh, I mean we've got 300 milliseconds in some cases and it's getting smaller, it's getting faster as we move to faster payments and even kind of cryptocurrencies as well, stable coins. But it's not going to be perfect all the time. Right. Such a small window, so much data. So how do we kind of go past or beyond real time kind of prevention and maybe even start looking at proactive? What would that even look like?

Speaker B: Yeah, I think when I mentioned before the ultimate outcome of a fraud scenario is that that financial loss, that like moment of a transaction and we have, we've really spent a lot of time trying to prevent that transaction from happening but often we're we're at that time already, which in a lot of cases is, is, is, is a little bit too late. So whenever we can find a way to prevent a transaction from even being initiated, then we're in a much better place. It's uh, you know, the, the effort that goes into recovering a fraud, recovering a payment, just the whole kind of management after the fact. Uh, if we can avoid all of that and get to the moment in time where the actual compromise is happening, then that's going to be way more beneficial to us. So we kind of think of that as, okay, so what's happening before a transaction? There's online activity that's happening. Can we identify those? Is it risky? Devices that are accessing the account, things like that. But then even before that, there's the point in time where the account's being opened, that account onboarding that account opening. Is there insights that we can, uh, pull from the network to say this, this individual is, um, risky based on other activity that's occurred across the network? Or then even, you know, kind of even before that. Are there other data sources? Are there other places that we can pull insights from? Thinking about like, there's a lot of different services out there that will gather data from the dark web or from these, like Telegram chats or other places where, where fraudsters are sharing information. So can we see that a check has been stolen from the mail and then this check was, is going to be used to, to, uh, to, you know, it's going to be sold at some point, so, or even pulling data across other sectors. Thinking about like social, uh, media, is there information and insights that we can gather there or you know, telcos, Is there, is there data across all those different areas that lead up to the point of a financial compromise, a transaction that can allow us to identify these scams earlier and prevent that whole, that whole transaction from, from even occurring? Yeah, there's real, like, you know, real time is key. There's those, those kind of last. It's really, it's your last opportunity to say this is risky, don't send this transaction. Um, but there's, there's so much that can happen prior to that transaction and so much data that can be leveraged, uh, to, to really allow investigators to, to see the, the risks before they end up becoming a transaction.

Speaker A: Yeah, it's a little bit reminds me of a Minority report, right? All that information beforehand. But when we're currently doing real time, there's only maybe five or seven valuable signals that can be used to kind of make that risk score or make that decision there and then. But prior there must be unlimited amounts of data points and signals. So how do we leverage these without it just becoming noise?

Speaker B: Yeah, that is a very good point. The, the one, one of the benefits of AI is the, the ability for um, you know, machine learning models, large language models, different types of AI to pick out the, the pieces of data that are beneficial and kind of take away all the noise. I think the uh, a lot of the traditional kind of rules based models are relying on uh, you know, somebody building a rule and saying this particular scenario, let's identify that, what's happening in that, that particular scenario. But a machine learning based model can gather in so many different data points and if you have enough labeled data to say hey this, this, in this scenario, this is fraud, then we can actually take away all those different things that are, that are noise and, and pull it into, to something that's beneficial to the investigators. The other piece then is just, you know, I think a lot of our customers talk about uh, the swivel chair problem or the um, just context switching problem where you're in a whole bunch of different systems and you're trying to gather data or piece together things in different systems. If you can pull all of those data signals into a single kind of unified view where everything is in one place and uh, all the information that investigator needs is right there to be able to uh, see the scenario, then uh, that's going to help not just the models but it also helps an investigator kind of contextualize what's happening in a particular scenario.

Speaker A: So there'll be no more 50 tabs open three different apps, uh, four different browsers.

Speaker B: I can't guarantee that there's not going to be 50 tabs. I think my minimum number of tabs is probably 50.

Speaker A: And I mean AI we've spoke about the traditional ML automation, generative AI does the whole kind of talk about agents and what agents and what their part plays in the future. So you talked about a little bit there. What, what else can we expect?

Speaker B: Yeah, so what we've seen, we've seen that agents, uh, agentic AI can really benefit fraud teams at uh, uh, financial institutions and really allow that uh, those teams to become much more efficient and essentially become super investigators on their own. So, so we actually, we launched our AgentIC AI workforce recently and have over 700 financial institutions that are using it now. And as with, with uh, our agentic AI workforce, we have essentially the ability for our AI agents to understand uh, the context of an alert, look at the scenario that's happening and provide recommendations on should this um, alert be investigated further or should uh, should we actually acknowledge it and you know, this is normal activity. Let's continue on. I think it's something that we've seen a need for in the industry, just that ability to expand the capabilities of an individual investigator. Uh, especially as we see that you know, criminals, fraudsters are continuing to leverage AI to perpetrate these frauds. As we mentioned before, there's more and more of these scenarios happening. So that kind of speed and scale that's occurring on the fraud side of things requires us on the defense side, uh, to, on the prevention side to make uh, sure that our teams, our investigators are as efficient as possible and they're, you know, they're not bogged down with those kind of regular resource intensive manual workflows and things that they, they, those manual steps that they have to do all the time, that piece can be automated by these agentic workers. Then it allows the, the investigators to focus their time on those, you know, those needles in the haystack that are the scenarios that are more challenging and require that, that human intervention.

Speaker A: And this is kind of cutting edge technology, right? Using agents in, in financial institutions in particular as well. So you've plus that are leveraging it so far good and bad. What would you think the biggest surprises have been? Putting these into banks who are uh, adopting.

Speaker B: Yeah, so the challenge like you said, when you're working with uh, any sort of cutting edge technology is having those customers who are willing to work with you to build it out and to see the value. I think what we've seen as more and more of our customers have adopted, um, the product is just more proof points on how effective they can be, how their kind of workflows change alongside the agentic workers there. So in many cases we're seeing the amount of time it takes to work alerts, even complex alerts, especially on the kind of AML side of things, uh, where those, those money laundering alerts can be really long, complex scenarios. The amount of time that it takes just significantly gets reduced because uh, not only are you providing that kind of recommendation and the information, uh, but then that can be used for documentation and you know, fit in with their, their existing processes or things like that. I think the one as we developed these agentic AI workers, one of the biggest benefits that we've seen is being able to not only leverage the. Again, I'll come back to the network piece. It's being able to Leverage the kind of consortium view the value of the network and pull those insights into uh, those recommendations. So then your agentic workers are not only kind of looking at the alert and determining what's what to do based on the alert, but they also have the benefit of the knowledge of the view across the network. So that kind of, that combination of, you know, very cutting edge AI plus a uh, network those two put together is really uh, kind of the what I see as the ultimate solution for effectiveness and efficiency in looking out for odd alerts.

Speaker A: Yeah, the secret sauce. It'd be interesting to look at what additional resources, technology we will need in the future. But right now it does appear that consortium data, as much data as possible and having the technology to filter, sort, organize, running alongside is probably one of the only options that we've got at the moment. Colin, it's been absolutely pleasure to talk to you today. You've got a great vantage point with so many different institutions and the data that you work with, with prediction markets are uh, popular at the moment. I call it gambling. But if you were a betting man within the fraud prevention industry, what would you be putting your money on is going to happen or what's going to change in the next 612 months?

Speaker B: 612 months? I would say more. It's probably going to be a relatively boring answer. More scams, more uh, but one I think we might start to see a little bit of a shift back to uh, some of the unauthorized scenarios. I think what I, you know, we've seen uh, you know, some of the really cutting edge AI models that are really good at finding cybersecurity threats or vulnerabilities in software. And I think my, what uh, I'm anticipating is that we'll see more, you know, more of those kind of stolen credentials, more scenarios where something has occurred on like a legacy piece of software and credentials get compromised somewhere where you wouldn't expect it. And then you get those kind of credential stuffing attacks where they're taking the information they've learned from a completely uh, separate piece of software. You know, your university password that you had 20 years ago and then it's also the same as your bank password. So that's because of the kind of change in pace of software vulnerabilities being identified. My, I'm curious to see over that next kind of 6 to 12 months if uh, if that's going to lead to more, more unauthorized scenarios.

Speaker A: Interesting. So everyone's obviously going, looking towards the authorized payments. So maybe it is that especially with the uh, OpenAI and hugging face issue as well. That's, uh, hopefully set, ah, a precedent and make sure that people don't forget about the cyber side as well. Absolute pleasure. If people wanted to learn more about the reports from the work that you're doing, where can they find you? Keep up with what's going on.

Speaker B: Yep. So the report's available on, uh, verifyn.com, i'm also on LinkedIn. Feel free to connect with me there and happy to chat with anybody on any of my favorite topics in the fraud space. Yep.

Speaker A: Awesome. Well, thank you very much for joining. Hopefully those, uh, predictions. I say hopefully. Let's see if those predictions come true. And we'll speak soon. And some listeners. Ciao for now. Want to hear more about fighting fraud with tech? Subscribe to Leading Detection wherever you get your podcasts.