Part 1: Automation to Detect and Stop Fraudulent Transactions

Part 1 Automation to Detect and Stop Fraudulent Transactions Informed

At the recent Bank Automation Summit, Informed’s Director of Auto Lending Strategy, Jessica Gonzalez participated in a panel discussion. Also on the panel:

Kevin Faragher, Senior Director of Product and Strategy at Ally Financial

Michael Reynolds, Business Technology Senior Manager for Service Digitization at KeyBank,and

Whitney McDonald, Deputy Editor Of  Bank Automation News (moderator)

The panel had a lively and engaging discussion sharing lots of great advice. There’s so much content that we’ll share tips and tidbits (edited for brevity) in a series of four posts. The final post has the answers to audience questions that weren’t answered live due to running out of time!

Whitney – Let’s begin with the influx of fraud that we’ve been seeing in the financial services industry. Just as more and more digital capabilities have launched, especially since the pandemic and lots of new technology. A good place to start, to put it into context, is what types of fraud are on the rise in financial services? And I know that Jessica has some figures to share with us.

Jessica – Rising fraud is a hot topic in large part because of the pandemic. Car buyers are using a digital interface to purchase and finance cars, so in auto lending, we’re seeing $4.7 billion losses. At Informed, we see an average fraud rate of 2.25% across all of our lenders. A digital presence actually increases fraud by .08%. What that means is that fraudsters are getting trickier, more sophisticated, and using digital platforms to enable them. 

We’re making sure that fraud is contained. Currently, law enforcement focuses on identity theft, because it’s easily punishable and considered a “hot crime.” So you see dealers and  law enforcement, really focused. We focus on paystub fraud because it helps ensure that consumers can pay back their loans. So instead of just focusing on identification or KYC, we calculate consumers’ income and are clear and transparent.

Whitney – You’ve talked about paystub fraud and I know you just released a bulletin on that. Can you talk a little more about what you’re seeing?

Jessica –  Sure. The fraud rate across all of our lenders and portfolios is about 2.25%. From a digital perspective, we’re actually able to find 35% more fraudulent activity. And if you’re a digital retailer, you’ll be 10x more likely to see fraudulent paystubs, and fraudulent documentation across lending. So when we’re looking at a lot of these trends, we’re comparing it against that average 2.25%.  You may think, oh, that’s not a big deal, but it’s worth billions of dollars. 

In addition, we see outliers. There are areas where segmenting by portfolio, dealer, FICO and non-FICO, you’re going to see trends that highlight the issue. So the key is not only having data to ensure you’re tracking fraud, but also making sure to recognize trends. 

As Kevin was saying, it’s difficult to manually track trends within a portfolio. Funders or analysts look at these documents and they’re seeing tons and tons of documents every day. They’re unable to combine those data points in a holistic way to understand the trends in what they’re seeing.

For example, when I was at the bank, we saw telephone bills. They had a different name, different address with the same telephone number as someone else. And it took us almost six months to identify that. So having real time, automated transaction analysis is imperative. You equip your fraud team who is looking at documents and also the industry by sharing data resources.

AI takes those millions of transactions and highlights fraud trends. So you have the data and ensure to use and analyzing it properly.

Whitney – I know that Jessica just gave us what she sees on the backside. Now, to the banks. We’ll start, with you, Kevin, with Ally. Can you share in terms of recent increases in fraudulent activity what you’ve been seeing?

Kevin – Sure. You think about how fraud used to be. It was some guy stealing somebody’s mail, then getting a fake ID and going to the dealership buy a car. If you were smart enough as an underwriter you recognize that this guy has a credit bureau note in California, and they’re applying for a car loan with us in Detroit, which didn’t make a lot of sense. But in today’s age, everything’s got to be fast. And speed is one of the integral business value propositions. 

I think that fits digital very well because fraudsters take advantage, trying to do things so much faster. One of the biggest types of fraud we’re seeing now is fraud where we have folks that are either partially or completely making up a credit profile that is designed to get through our underwriting systems. I recently saw an example where somebody had their credit score increased or improved with the model trade line which made the deal score better. 

So we review all the data we receive and we do a simulation. We have people looking at them manually, but they’re really hard to spot. When the deal comes through with a synthetic ID you still have to support the identity and when they provide fraudulent paystubs, that’s where we see a big pay stub and having the ability to have the AI capture that and flag it for our people is really valuable. 

And Michael, can you talk about fraud that’s on the rise at your bank?

Michael – Yeah, definitely a couple areas. Anybody a Netflix fan? If so, you’ve probably seen Ozark. So you can be like Marty Bird and figure out how to create 14 accounts. Then you move some funds and they gradually grow and he teaches his kid how to do it. Well, fraudsters are much more advanced than that but that’s a simple example. So think of it this way.

There was a case about nine years ago, where all of the company’s contractors’ information, including fingerprints bled out across the dark web. Now you have enough information to open an account based on an existing person’s current banking relationship. And then switch that up just a little bit where now that existing account has maybe $10,000.  The account owner doesn’t have bad credit on this account so now the fraudster hits your maximum limits and the bank will cover that for the account.

So now we go back to Marty Bird who’s got all these different accounts. I hit your maximum overdraft limit coverage and now I’m generating funds and getting shifted into up to 14 different accounts affecting many people’s lives as well as many banks. So the sophistication and volume are definitely trending up. And remember I’m the automation guy,, so I see a lot  arising and all of a sudden now I gotta get another virtual machine and manage the volume we’re seeing.

Whitney – Great, thank you. And now, Kevin and Jessica, you both started talking about this. How Ally works with informed.IQ in terms of flagging real time fraudulent transactions. 

Jessica, can you talk through how banks can leverage this technology –  maybe give us a how to guide?

Jessica – Informed automatically detects fraud on paystubs, which is one of the first fraud entry points into the lending process, because it is so easy to get a fake paystub. So it’s imperative to understand that we think of fraudsters as really high tech and while that can be true, it is also everyday people faced with a barrier to entry. As Kevin stated earlier, if you only focus on non- documentary verifications you might run into a large amount of synthetic IDs.  If you focus on KYC and identity fraud but do not consume digital documents there is a limitation on how much automatic detection you can enable. 

It starts with the digital experience, from the infrastructure perspective, because if you’ve received a flat image from a consumer, just a document image from an email or fax, image quality can be an issue. So if you get a fax or a picture of a picture, it’s very difficult to know if that identification is fraudulent. AI can focus on the ID, but if it’s a flat picture you’ll only find 10 to 20% of the time the AI can determine whether or not it’s fraudulent.

Most of the lending industry is still highly reliant on paper documentation so Informed is meeting our industry where it can make a significant impact – where we have high confidence that we are uncovering fraud. So relying on Informed’s paystub fraud measure is a good indicator for lenders to make sure they’re able to identify not just KYC but more of enhanced fraud. Maybe somebody looking at it is not able to see it and because it is a lot easier to get a fake paystub than a fake ID and there is more focus on KYC and ID verification, more paystub fraud is likely to occur. 

I also think that making sure lenders have the ability to do account opening or have a seamless experience where you’re able to have consumers upload a document is really important. Because if you have those checks at the very front end, you’re able to reduce your fraud very significantly. So making sure that you have at the beginning of your waterfall enabling for that first broad check is critical. We actually see that poor image quality actually resonates and correlates to poor performance within the loan portfolio.

And I have a little sneaky suspicion that it’s because of the fraud aspect there. If you have people that can and will repay loans but cannot provide supporting documentation they most likely will go towards synthetic ID or a CPN but when we see pay stub fraud they are more likely to go into default. They don’t have the means to make those payments.

We’ll wrap this episode here. You can read the second episode here. Following the 3-part series we’ll post a bonus episode answering all of the audience questions that came in following the panel. If you like to reduce your fraud levels contact us to review your process.

author avatar
We are digital transformers of tedious manual tasks, bringing robotics process automation to financial services.

New: American Banker - Sharing information = best defense against AI fraud