5 Things to Know About Detecting Credit Card Fraud
MikeEbbers 270001G1G2 Visits (25558)
Did you ever wonder why the owner of a stolen credit card is not charged for fraudulent transactions? One reason is because fraud can quickly be detected with a computer by tracking usage patterns and history. IBM has solutions to help. Here are 5 things to know about IBM fraud detection solutions.
1. Companies get ripped off by billions of dollars each year due to fraud.
In the U.S., 32 p
2. The top 25 world banks run their businesses on mainframes.
In fact, 71% of Fortune 500 banks use mainframes. These facts are seldom publicized, but should be no surprise. IBM System z mainframes have experienced nearly 50 years of improved hardware, software, and procedures, making them reliable and quite foolproof. You don’t often (if ever) hear of someone hacking a mainframe.
3. The ideal solution avoids making fraudulent payments without slowing down legitimate payments.
Such a solution requires the adoption of a comprehensive fraud business architecture that applies advanced predictive analytics to reduce fraud, waste, and abuse, by using the following techniques:
IBM fraud analysis and scoring models are supported by such familiar products as System z, the IBM DB2 Analytics Accelerator (IDAA), DB2 for z/OS, and SPSS. New functions in DB2 for zOS and SPSS Modeler allow the solution to be run in close proximity to the vast amounts of historical data that is on System z. This allows the scoring of a payment to be made directly within DB2. Figure 1 shows the layout and flow of this solution.
Figure 1. IBM real-time fraud detection solution on System z
4. IBM has created an SPSS scoring adapter that runs in DB2.
The scoring process uses live transaction data as the input and produces real-time results. Figure 2 shows a sample SPSS Modeler stream that gets published to the scoring adapter.
Figure 2. The SPSS scoring adapter in DB2 is the heart of the solution
5. The brains behind predicting scoring ratings is a user-written SPSS model.
For a typical transactional fraud detection business case, assume that a customer is making a credit card payment. At the time of payment, the bank analyzes the payment pattern on that particular credit card to detect the possibility of fraud. This analysis involves the history, frequency, and dollar amounts of previous transactions for that credit card from its database records. Depending on the scoring analysis, the bank authorizes the transaction, keeps it on hold, or declines it, all in real time.
Figure 3. The SPSS model is the brains of the solution.
Don’t you love a happy ending? Having a credit card fraud detection application can provide one happy ending after another. And this application can be applied to other types of fraud detection including insurance. If you would like to read more, see Real-time Fraud Detection Analytics on IBM System z, SG24-8066.
Mike Ebbers is an IBM Redbooks Project Leader. He works with technical experts to create books, guides, blogs, and videos. Follow Mike on Twitter at @MikeEbbers.