In this series of posts, I will develop different ideas about how fraud in e-commerce, and different techniques which can be used to counteract this type of behavior. Some of them are standard knowledge acquired at my time working in fraud detection, other are from the corpus of information that has been developed in the industry, also a couple of my own ideas which I would like to have implemented in the future.
Sources of fraud
Fraud in e-commerce has mainly a two sided profile : Payment fraud or product fraud. Either the money one applies to an operation is lost, or the received product represents a lesser valued than payed (Which is another way of lost money)
Credit card / payments processor fraud
Credit cards are nowadays the normal payment type of a online transaction, be it directly or indirectly through a payment processor.
Fraud types in payments have many ways of being perpetrated, but in this simple article, we will generalize and leave apart some of the more sophisticated, to bring a very general idea of the modus operandi on each regard.
Rush orders
This phenomenon occurs when a fraudster has access to credit cards credentials, and so he/she makes rush and overnight orders, to try to snap all possible credit he can before the card is reported.
The quickest shipping methods are also employed for this type of fraud, because once the product is on the hand of the buyer, the order cancellation won't prevent the goal of the operation, at least the primary one.
Product fraud
This source of fraud is related to the product owner "making believe" the buyer of the product quality, but in reality the intention is never to release the product. For this type of fraud, very attractive prices are set, and they are sold at a very high speed, so the user has a very short span of time between seeing, thinking, and buying the good. It also can be combined with the following technique, requiring the buyer to pay by means outside the control of the site.
Site mechanisms fraud
This type is the most frequent fraud type in online auction sites. The objective of this type of operation is sending the buyer information embedded in the product's media: Descriptions, photos, comments.
There are really interesting techniques in this regards, including optical character recognition of text patterns in the images, simple text pattern recognition, and even neural networks applying Natural Language Processing, and automatically getting text patterns associated with embedded information sharing, never before though from the analyst, learning from a list of positive previous cases.
Summary
In this first post, I've set a very minimum structure to develop different fraud detection techniques divided in a couple of main fields. Many other classification types can be employed, please let me know in the comments how this classification can be enriched and completed.
