Investigative Data Mining for Security and Criminal Detection

8.7 Common-Sense Rules

Before an e-business can spot and stop on-line fraud, however, it has to know what to look for. This commonly starts by developing a set of rules describing a fraud profile. A fraud profile summarizes the data characteristics that one would expect to find in questionable transactions. There are several common-sense rules that experienced fraud specialists look for in detecting and deterring this type of crime in the terrestrial business environment. They include orders with the following red-flag conditions:

• Multiple or single orders that fall just under the "review threshold" level

• Shipping addresses matching current or former employee's addresses

• All orders with different shipping and billing addresses

• Any returns, rejects, and for-credit orders

• All P.O. box address orders

These fraud rules can be coupled with models created with data mining software to detect and deter on-line fraud by e-businesses. As in the past, these types of tools gave merchants the ability to search quickly through millions of records in a matter of seconds in order to identify transactions in real or near real time that have the characteristics associated with fraudulent activity. Through the development of these common-sense rules and the use of predictive models created with data mining tools, merchants have the ability to reduce their losses and double-check certain orders before shipping them out.