US 6,330,546 · Granted 2001-12-11

The Neural Network That Catches Credit Card Fraud Before You Notice

Imagine a security guard who learns what normal shopping looks like for you—then spots when something weird happens and blocks it. This patent describes exactly that: an AI system that watches your account, learns your patterns, and catches fraudulent transactions by comparing new activity against what it knows about you.

Curated and verified by the IsItPatented Editorial team

The plain-English version

What it protects

The claim covers an automated fraud-detection system that uses a neural network (a type of machine learning) to build a profile of individual customer transaction behavior, then identifies potentially fraudulent transactions by comparing new activity against those learned patterns. What's protected here is both the method of training the model on known customer data and the system's ability to generate reason codes—explanations for why it flagged a particular transaction as suspicious.

Why it matters

Before machine learning became mainstream, fraud detection relied on simple rules that criminals could easily work around. This patent represents an early commercial application of neural networks to real-time financial security, allowing banks and payment processors to catch fraud faster and with fewer false alarms than rule-based systems. The assignee, HNC Software, pioneered this approach in the late 1990s during the early e-commerce boom when online fraud was a genuine barrier to customer trust.

Real-world use

Every time your bank declines a suspicious transaction—like someone trying to buy plane tickets across the country in an hour—that decision likely traces back to predictive models descended from systems like this one.

Original USPTO abstract

An automated system and method detects fraudulent transactions using a predictive model such as a neural network to evaluate individual customer accounts and identify potentially fraudulent transactions based on learned relationships among known variables. The system may also output reason codes indicating relative contributions of various variables to a particular result. The system periodically monitors its performance and redevelops the model when performance drops below a predetermined level.

Patent details

Publication number: US 6,330,546
Filing date: 1998-10-05
Grant date: 2001-12-11
Assignee: Hnc Software, Inc.
Inventor(s): GOPINATHAN KRISHNA M., JOST ALLEN, BIAFORE LOUIS S., FERGUSON WILLIAM M., LAZARUS MICHAEL A., PATHRIA ANU K.
CPC class: G07F7/08

View original record on Google Patents →

Want to file your own patent?

If you're building payment or security software, understanding how machine learning models protect financial data is essential; explore similar fraud-detection patents on IsItPatented to see how the field evolved.

Free patentability scan