Rethinking Your Fraud Approach: Network Analytics
Recently we explored the “COVID-19 fraud effect”—that is, how the new shelter-in-place orders to curb the spread of the virus has boosted time people spend online. With more consumer time online comes a swift rise in opportunities for fraudsters to attack and steal PII. That means businesses must be ever more vigilant in how they approach fraud detection and prevention. Hence, we launched this series on “Rethinking Your Fraud Approach” which will explore tactics you can use to mitigate fraud against your business and your customers.
The first tactic—the application of network analysis — starts with data you’ve already collected about your customers. Network analysis is a technique used to evaluate the relationships (or connections) between data, such as customer ID, email or physical address, payment methods, past orders and so on. The goal is to identify abnormalities that may indicate fraudulent activity.
Here are some common examples of relationships to evaluate:
- Distance to known fraudsters
A new customer applies with a phone number that matches the phone number of an existing customer who was found to be fraudulent. The new customer may be fraudulent, having only one degree of separation from known fraud.
- Number of associations
A second customer applies with a device fingerprint that matches the device fingerprint of a known fraudulent customer and an email address that matches the email address of another known fraudulent customer. This new customer is more likely fraudulent than the first new customer, being associated with 2 known cases of fraud instead of one.
A customer typically logs into their account with device fingerprint X. Today, the customer logged in with both device fingerprints Y and Z. This behavior alone may not be indicative of fraud, but when combined with “distance to known fraud” and “number of associations,” it could be.
Data relationships such as these can be defined, monitored, and stored on the backend as link variables. By creating alerts when the variables reach a certain threshold, you can detect activities that point to account takeovers, synthetic identities, and other fraud techniques.
To block fraud activity before it can inflict significant financial and reputational damage, you’ll want to automate and optimize the back-end processes as much as possible using machine learning techniques and a real-time decision management service. We’ll talk more about how to do that in an upcoming post in this series.
At Enova, we’ve had 15+ years in successfully detecting and preventing fraud across our many online consumer and small business lending brands and Enova Decisions has packaged much of our know-how and technology into solutions that let you reduce fraud losses by up to 75% in the first few months.
Stay tuned for our next post exploring Velocity Variables or contact us today to get better fraud detection and prevention up-and-running quickly.