Credit card fraud is a currently a widespread crime. In fact, in 2014, out of the 17.6 million identity theft cases that were filed with law enforcement agencies, 86% of all the victims filed fraud reports linked to their existing bank account or credit card.
Also, the Federal Trade Commission revealed that credit card fraud is currently the most widespread type of identity theft in the United States, with over 130,000 reports of such cases each year.
Although automated techniques of identifying suspicious patterns of credit card usage are not new, eBay researchers recently described an innovative method in a paper dubbed (“Credit Card Fraud Detection in e-Commerce: An Outlier Detection Approach“), which was published in Arxiv.org.
Their recommended system utilizes an algorithm that is trained to identify “good behavior,” as it associates with both payments and transactions, as well as to flag any activity that does not correspond with the expected norm.
“Often the challenge associated with tasks like fraud and spam detection is the lack of all likely patterns needed to train suitable supervised learning models,” the paper’s authors wrote.
“This problem accentuates when the fraudulent patterns are not only scarce, they also change over time … Limited data and continuously changing patterns makes learning significantly difficult. We hypothesize that good behavior does not change with time and data points representing good behavior have consistent spatial signature under different groupings.”
The researchers used an “ensemble” of grouping techniques, which entail methods that are used in identifying sets of similar objects in a given dataset, with varying parameters.
Each data point was allocated to a given group in each training.
From each of the training run, a mathematical representation was generated, comprising “fingerprints”, particularly those of the data point that could be incorporated into its distinctive signature representation.
To produce a signature that showed “good behavior”, the team integrated the per-data point mathematical representations and, in turn, weighed them based on the size of the particular cluster.
They arrived at one score, particularly between 0 and 1.Low consistency, which is a score near 0, naturally matched with outlier behavior.
The authors wrote that the technique had some advantages over the detection of traditional AI fraud.
It did not call for prior knowledge of inliers or outliers.
Also, the basic algorithm was both general in nature and highly scalable, which means that it could be used in virtually any grouping problems such as those found in medical domains.
If you have bought or sold an item on eBay of late, you may have come across this system in action.
The researchers also modestly pointed out that it was successful in selecting fraudulent transactions, particularly in the data drawn from an “e-commerce platform”
“The motivation for [our] approach comes from trying to identify fraudulent consumers on an e-commerce platform … Each time the e-commerce company introduces new consumer aided features or imposes restrictions on certain transactional behaviors, it opens new doors and avenues for some consumers to misuse and abuse the platform. Our algorithm shows tremendous potential in identifying [fraud] … However, due to the confidentiality of the dataset, these results cannot be reported in this paper.”