Applying Cross-Channel Data to Improve Fraud Detection

Feb 19, 2018 | by Heidi Bleau

In light of industry initiatives and global regulations – including PSD2, 3D Secure 2.0 and Faster Payments – a single question remains open:  Does risk-based authentication really work?  This is usually followed by inquiries on fraud detection rates, false positives and the impact on customers.  The benefits of an approach based on machine learning has proven to be extremely effective by consistently demonstrating fraud detection rates up to 97%. 

Despite these results, is it possible to predict fraud even more accurately and minimize the number of good customers who are challenged?

The Ecosystem Approach —Unique and Powerful
Fraud risk management has become a burden in recent years, and not just because the attackers have gotten better at their game. The tools and technologies used to detect and mitigate fraud events are better, but they are also plentiful. A survey commissioned by RSA found that 57% of organizations use between 4 - 10 different tools within their anti-fraud operations.

In addition, the face of digital banking and commerce is changing rapidly.  There are new channels, new interactions and new types of payment methods that customers are using to engage with organizations which only adds to the complexity.

Leveraging machine learning, organizations are able to use an ecosystem of data including third-party risk information from other fraud prevention tools as well as their own business intelligence to enhance risk scoring models and improve fraud detection.

A Case Study

A case study with a major U.S.-based financial institution illustrates the benefits of a data ecosystem approach. The bank had gathered extensive knowledge of risk factors from its call center and wanted to leverage that data to enhance protection within its digital channels. Using six custom facts supplied by the bank and applying them to the riskiest transactions (the top 1 percent), the client realized the following improvements to their existing fraud detection rates:

  • 2% increase in existing fraud detection rates across all event types
  • 3.6% increase in fraud detection for login transactions
  • 1% increase in fraud detection for payment transactions

Even just a 1% increase in fraud detection for payment transactions yielded significant results amounting to an average monthly fraud savings of $44K — or more than half a million dollars in a year.

# # #

Ready to See Results for Yourself?
Business-generated intelligence and risk data from other fraud prevention tools, when combined with machine learning models, can open up a world of possibilities for your fraud risk management strategy. Watch this short video on how you can apply cross-channel data to predict fraud risk more accurately. 

Author: Heidi Bleau

Category: RSA Fundamentals

Keywords: Fraud Detection, Fraud Prevention, Machine Learning, Mobile Fraud, Omnichannel, Online Fraud, Risk-Based Authentication, Web Fraud