Current Fraud Detection Methods Create too many False Positives

There is a flaw in fraud detection methods employed by the payment industry today: too many false positives. How many transactions pop up as potentially fraudulent, receive human inspection, and may receive human rejection, without any proof whatsoever of a consumer intending to commit fraud.

On the other hand many fraudsters learn that by creating transactional data that does not set off normalcy meters, their  fraudulent transactions receive approval without going anywhere near a human eyeball. Yet there are elements within message that shows prima facie fraud, meaning you could bring the message into court and just the message, and gain a conviction of fraud from the consumer that caused the origination of the message.  The difficulty with the approach is the haphazard collection and discard of data that occurs while a transaction message is in-flight.

For example, say a store operates at an address and a purchase message arrives from a point of sale (POS). The acquiring node tossed the address associated with the phone number once the needle created a circuit; however that address may have nothing to do with the POS operational address. If the two addresses are not the same, then it is possible to say fraud exists because a party to the transaction created the message from a place not known to the authorizer.   

I define machine examination of disparate data maintained in the entire set of transactional data (not just data associated with a part of the journey) as a filter. Millions of potential filters exist but fraud detection engines do not use any of them, because the prejudice to use deviations from known behavior, or newness of a card initiator, waste enough time, without tracking and pulling transactions that contain data that confirms fraud without false positives.

The fraud detection filters have a companion. Evasion detection filters capture transactions deliberately created to evade fraud detection filters and the two filters together create a dialectic response to the cycle of criminal activity. They also show when filters fail because they are too well known by the attacker class.

Next Blog: Find out tomorrow, or response to comments, or is there a demand for payment escrow accounts?