Profile of an online fraudster
I recently read a study about the profile of a cybercriminal. While I appreciate the study itself, one thing it lacks perspective on is an understanding of how identity data is being used to perpetrate fraud in the online channel. One may jump to conclusions about what is a good indicator for catching fraudsters. These very broad-brush observations may result in an overwhelming number of false positives without digging in deeper.
Purchase value
A single approach for understanding the correlation between purchase value and fraud does not work to best protect all businesses. Back in 2005, we saw that orders under $5 were great indicators of subsequent large-ticket fraud. For merchants that sell large-ticket items, such as electronics, those same rules may not be effective. To simply believe that the low dollar amount is the extent of the crime and not just a precursor to the real, bigger crime indicates a lack of understanding of how fraudsters work to manipulate a system. For some merchants, where fraudsters know they can go to do card testing against their business, low-dollar-amount rules may apply. However, for other businesses a different set of rules must be put into place.
Time of day
We have been tracking fraud time of day as a rule since 2004, but the critical point is a clear definition of which time of day. For the merchant, 3 a.m. is very different than 3 a.m. for a fraudster who is in Asia or Eastern Europe, where 3 a.m. merchant time is actually the middle of the online fraudster’s day. FraudNet is designed to identify the time from the user’s device and runs its rules from the user’s time.
We find that every individual business will have a very specific threat profile. Businesses need to build their individual fraud strategy around their overall attack rate taking into account the strength of the defense and the ability to be flexible to accommodate the nuances for individual consumers. A general approach to fraud mitigation inevitably results in a system that begins to chase broad averages, which leads to excessive false positives and mediocre detection. That’s what drives us to do the job better.
The proof of every fraud solution should lie in its ability to catch the most fraud without negatively impacting good customers.
