Feedzai Fraud Detection Platform Uses AI, Biometrics

If you’re preparing to withdraw money at an ATM in a foreign country, take a good look at where you insert your card. Go ahead, grab it and give it a little shake. The plastic case your card goes into is sometimes replaced by a phony case which captures your pin and card info while you’re conducting a legitimate transaction. When you’re walking away from the ATM, don’t react to any “Watch Out For Pickpockets!” signs you might see. Police in Hungary discovered that pickpockets themselves were putting up the signs. That’s because people instinctively reach for their valuables when warned of pickpockets, immediately telegraphing where da loot at.

If you think there’s lots of fraud happening in the real world, wait until you see what’s happening in the digital world. That’s probably why Feedzai, another private company to recently join the unicorn club, is gaining traction for its machine learning platform to prevent and detect fraud and money laundering in the financial services sector.

Fraud is on the Rise

And if you believe the statistics – many of them from Feedzai, which issues quarterly financial crime reports – fraud in the financial services sector is out of control. All banking fraud across the Internet, telephone, and in-person banking (people still do that?) – grew 159% in the first quarter of this year compared to the last quarter of 2020. Online banking, which accounts for 96% of all banking transactions, made up 93% of all fraud cases. The analysis reportedly covers more than 12 billion global banking transactions. 

But let’s remove the implied bias and turn to a neutral party: The consumer credit reporting behemoth TransUnion, comparing a similar period of time, said that suspected financial services digital fraud attempts increased 149%. That’s what we could call a growth market:

Growth of the fraud detection and prevention market.
Growth of the fraud detection and prevention market in billions. Credit: Statista

With a company name that’s impossible to spell over the phone, Feedzai seems to be putting itself in the pole position within the fraud detection and prevention market.

A New AI Fintech Unicorn

Click for company website

Founded in 2009, the Silicon Valley-based startup has raised $277.5 million in disclosed funding, with most of the money coming from a massive $200 million Series D in March. That vaulted the company to a valuation of $1 billion.

While most of the names on the investor list skew toward the venture capital (VC) side of the spectrum, such as Sapphire, another notable name is the venture arm of Citigroup (C). It’s usually a pretty good sign when one of the largest banks in the world not only buys your product but also invests in it, including an undisclosed venture round back in 2016. In fact, Feedzai’s war chest may be north of $300 million, as the private company just raised another undisclosed venture round at the beginning of this month from a little VC firm based in Madrid, Spain called Conexo Ventures. The investment will make a little more sense later on in our story.

How Does Feedzai Use AI to Detect Fraud?

The answer: Pretty damn quick. The startup claims its platform can calculate a risk score in three milliseconds, evaluating thousands of decisions to score a transaction in real time. Machine learning models identify patterns in the transaction by analyzing customer behaviors and profiles, as well as leveraging external, third-party data that the models use to assess the risk of the transaction and detect financial crime patterns. Customers can customize the models by adding other rules and lists to fine tune the decision to approve, reject or review. The result is reportedly transparent with clear documentation for the decision.

Feedzai AI platform.
The Feedzai AI platform clearly documents its decisions and provides cool-looking visualization tools to detect larger patterns of fraud. Credit: Feedzai

But it’s not all up to the machine: A risk analyst reviews the transaction scores and fraud alerts, passing on anything suspicious to a company’s senior data analyst. This person uses Feedzai’s AI-powered visual analysis tool called Genome, which can help uncover larger patterns of financial crime and fraud. All decisions are fed back into the machine learning model to improve performance over time. 

Feedzai Use Cases

The platform protects financial institutions against funny business involving account openings, transactions, and money laundering. Its clients include retail banks, payment processors, and merchants, among others. Feedzai claims it monitors companies that serve more than 800 million customers in 190 countries, with products protecting half of the populations in the UK and Canada, as well as four of the five largest banks in North America. But here’s the stat that really caught our attention: More than 20% of the world’s money flows through the Feedzai platform, according to the company.

Feedzai customer benefit statistics.
Credit: Feedzai

In addition to Citibank, other large financial institutions such as Fiserv (FISV) and Santander (SAN) are also customers, along with fintech companies like SoFi (SOFI) and Mox by Standard Chartered. Another customer that is part of the new generation of financial institutions is SafetyPay, a digital alternative payment platform focused in Latin America and Europe. SafetyPay allows consumers to shop online via bank transfer or cash without sharing their information online.  Feedzai comes into the picture by monitoring patterns in payment activity and comparing them against a customer’s historical data to authenticate transactions. 

Feedzai Acquires Revelock

Click for company website

No doubt Feedzai will take some of its new fortune to further build out its platform. In fact, the company wasted little time doing just that by acquiring Madrid-based Revelock, a behavioral biometric platform, for an undisclosed amount of money last month. Founded in 2015, the startup had raised $14.3 million to date. Most of the investors are headquartered in Spain, including a name that showed up earlier – Conexo.

Biometrics employs various technologies to verify or authenticate identity, including everything from facial recognition on your iPhone to voice verification to the way you handle your mobile device. In the case of Revelock, the company has developed a hybrid AI model that powers BionicID, a unique digital identifier that is created by collecting each user’s thousands of biometric, behavioral, device, and network data points. It continuously analyzes user behavior, from login to logout, across every interaction. If everything appears cool, the session continues uninterrupted. Otherwise, Revelock goes into action, which might involve verifying the user with stepped-up authentication, terminating the user session, or locking the account.

Revelock BionicID platform
Apparently ticks are part of the threat intel. Credit: Revelock

Feedzai has some grand plans around this acquisition to create the “world’s largest Financial Intelligence Network, a vault of more than a trillion data points, sessions, and profiles of both good and bad actors.” The network would combine Feedzai’s 150 billion data points for fraud detection with Revelock’s BionicID, a sort of human digital twin, based on how it’s described. Of course, every session promises complete anonymity to “ensure people are treated as people and not as data points.”


The digital shift of the economy is becoming seismic, with The Rona shaking things up even harder when people started treating paper money like the plague, with ATM withdrawals down 58% in 2020. About 41% of consumers switched from cash to online/phone payments and more than half of them do not plan to switch back to cash, according to Zelle, the mobile payment app that defaults to sharing your every transaction on social media. Meanwhile, the bad guys haven’t missed a beat, finding new ways to separate people from their money and identities. Fintech companies like Feedzai are betting they can outsmart them with artificial intelligence. 


Leave a Reply

Your email address will not be published.