Hadoop Application Architectures

Chapter 9. Fraud Detection

What is fraud detection? Well, yes, it’s a multibillion-dollar industry that touches every company with significant amounts of money to lose, but what is fraud detection at its heart? For the purpose of our discussion, we can view fraud detection as making a decision based on whether an actor (human or machine) is behaving as it should. There are two different points here: 1) knowing the difference between normal and abnormal behavior, and 2) being able to act on that knowledge.

An easy example to illustrate the first point is parents knowing when their child is telling a lie or hiding something. If you are a parent, you know how it is: you see your child every day and you know his normal behaviors and patterns. If one day he’s a little quieter than normal or trying to avoid eye contact, your instincts kick in and you start asking questions until you find out he failed a test, got into a fight with a friend, or got bullied at school. You were able to detect the deception because of your close relationship and knowledge of your child. That close relationship is key to detecting changes in well-established patterns.

Continuous Improvement

Now, our example will rarely end in your child stealing millions from a bank or getting involved in money laundering, but, like the actors in those examples, the child is trying to hide something. Just as your child becomes better at deception as he grows older, scammers and hackers become more sophisticated as time ...