Neural Network Based Fraud Detection System

By Joyce Moed, Reporter

HOUSTON–PULSE launched its new neural network based fraud detection system across its entire network last week, following a three-month beta development program involving 14 financial institutions. DebitProtect provides risk-scoring for all PULSE PIN, signature and PINless debit transactions on the PULSE network, and notifies designated financial institution personnel via e-mail of possible fraudulent transactions.

“Because debit cards are used differently than credit cards, PULSE designed the DebitProtect system based on a highly customized data model developed specifically for debit portfolios,” said Dave Schneider, PULSE president.

Schneider said that the service was developed for ease of use, and does not require any technical changes on the part of PULSE participants.

“It’s very user-friendly,” Schneider said. “All types of financial institutions can use this.”

DebitProtect is the latest addition to the suite of anti-fraud solutions PULSE offers its more than 4,400 financial institution participants, of which one-third are credit unions, Schneider said.

Debit transactions in the United States have tripled since 2000, Schneider said, and debit card fraud has increased. U.S. financial institutions lost about $662 million to debit card fraud in 2005, he said.

That’s why DebitProtect is so important, he said. DebitProtect uses technology to evaluate key transaction characteristics against a customized and constantly evolving risk model built from actual transaction date. The DebitProtect system is able to “learn” fraudulent behaviors and identify suspicious transactions that either are outside of established cardholder patterns or match fraud behaviors.

Risk scores are derived from card usage patterns, pre-defined modeling parameters and data from the transactions themselves, Schneider said. Scores that exceed a preset threshold trigger alert e-mails that are sent to issuing institutions. System users then log on a secure website to review information about the suspect transaction, classify it as either fraud or not fraud, and determine the next course of action.

“Security around PIN debit transactions is a very important subject,” Schneider said. “Enhancing and preserving that security is something we take seriously. We felt we were in a position to develop something particularly valuable. In the recent surveys with our financial institutions, fraud was recognized as the most important issue.”

PULSE’s DebitProtect system is being implemented in two stages: The phase 1 rollout provides PULSE-participating debit card issues with near-real-time transaction monitoring and e-mail notifications 24/7. When phase 2 is implement later this year, PULSE plans to support automatic blocking of transactions based on rules and parameters established by the issuer.

Schneider explained that this program is very much focused on debit cards, and not on credit cards. “Debit is more local, with smaller tickets and has a greater percentage of ATM transactions,” Schneider said, noting the main differences between debit and credit card usage.
Perhaps one of the most important features of DebitProtect, Schneider said, is its speed.