The holiday season and year-end months account for as much as a fifth or more of annual sales for some retailers, with 2023 set to be no exception. Thanksgiving and Black Friday may have passed, but as consumers get ready to enjoy the final days of the holiday season, it’s the perfect opportunity for fraudsters to attack. Research from TransUnion found the average number of suspected digital fraud attempts on any given day between Thanksgiving and Cyber Monday 2022 was 127 percent greater than during the rest of the year for transactions originating in the United States.
New research by Au10tix has revealed organized crime within North America is on the rise — the report finds a 44 percent increase in Q2 of 2023. Combined with the growing number of payments channels, and the rise of the omnichannel experience — where retailers have been forced to adapt to upwards of 10-15 different payment methods in-store and online — pressure is on to ensure consumer and reputational safety, while mitigating criminal activity.
Caught in the Act
In the real-time world of today, retailers need a real-time solution. Enter the new generation of event-driven architecture (EDA) that can help address growing fraud threats through detection, action and staying ahead of the rest.
As more retail organizations turn to data modeling, artificial intelligence and machine learning to recognize questionable transactions, EDA can take fraud management one step further. The challenge is to feed transaction data in real time to AI/ML, and it’s where EDA can provide real-time integration to communicate with modern micro-service payment frameworks and cloud-based AI/ML for fraud. Then, EDA and the event mesh allow flexibility so platforms can evolve and react quickly to changes.
First Things First — Build a Model
A fraud prevention model uses a set of triggers, which the data must be then fed into the model and assigned a score. The score also depends on authentication requests. If the user has reached the same score, yet there’s no biometric data or mobile authentication, then this would be highly likely to trigger a different reaction — blocking or flagging the questionable transaction for escalation.
Adding AI and ML to the Mix
It’s time to add AI and ML, which to be fully effective need a big data set as decisions are based on historic datasets. The first action is to “train” the model. Then the model runs through several fraudulent transactions so it’s trained on what a fraudulent transaction looks like so it can understand and pick out the right activities to be flagged.
EDA on the Lookout for Fraud
Layering EDA on top allows retailers to build an enterprise IT architecture that lets information flow between applications, microservices and connected devices in a real-time manner as events occur throughout the business.
Once the fraud model has been implemented across all transactions and payment channels, EDA enables the organization to leverage its fraud model and use AI/ML technology in real time across an ever-expanding number of payment channels.
Introducing the Event Broker
A retailer has to make sure payment channels just send the right event to communicate with the fraud detection system and receives the same events to get the “yes or no” back. Enter the middleman, the event broker. Event brokers enable what’s called loose coupling of applications. This is essential because it means applications and devices don’t need to know where they’re sending information or where information they’re consuming originates. But the event broker does.
AI, ML and EDA: The Perfect Trio?
With EDA, retailers can accurately support a high volume of transactions in the quickest response time. They can balance transaction authentication and authorization with fraud detection without decreasing customer satisfaction, and route events securely across the whole payments ecosystem with efficiency.
Ush Shukla is a distinguished engineer at Solace, an enabler of event-driven architecture for real-time enterprises.
View Original Article
