Share this post:
According to the 2019 AFP Payments Fraud & Control Survey, Automatic Clearing House (ACH) payment fraud increased significantly in 2018, reaching a new record with 82% of organizations reported incidents. Specifically, the percentage of companies that encountered ACH credit fraud jumped to 20% (from 13%), and those who experienced ACH debit fraud rose to 33% (from 28%). To further compound the problem, 25% of companies haven’t gotten advice from their banks on mitigating ACH risks.
Single channel-focused detection systems are easily fooled
Current fraud detection systems are designed to be channel-specific or focused on a single channel that can be too easily manipulated by criminal. These fraudsters can conduct part of a coordinated fraud scheme in each of several channels, never showing enough of a complete fraud scheme to be detected or even suspected in any one channel.
Figure 1: Multi-channel fraud schemes are not easily detected by siloed fraud prevention solutions
To see how this works in real life, see the example above.
- A compromised online channel creates a fraudulent overdraft line of credit, but no suspicious withdrawal.
- Shortly after, three ACH push transactions overpay three established credit card accounts through existing bill pay links.
- Two of the overpaid card issuers for Cards A and B send refunds to the compromised account and the third issuer for Card C posts a credit balance.
- Shortly later, the refunds are withdrawn as cash, and the credit balance on Card C is used to buy a high ticket item like a big-screen TV, which is then quickly added to a shipping container on its way to Eastern Europe.
The online channel saw nothing suspicious: a new line used to pay down card accounts is routine; nor are refunds received and withdrawn inherently suspicious. The card accounts look like innocent overpayments and the TV purchase involves no draw on the card credit line — what’s to suspect? Only by seeing the details together, across the channels will such a fraud be detected. And once such a scheme is perfected, it will be repeated hundreds of times. But, a custom model built on integrated multi-channel data would identify and flag a pattern like this in short order. Most fraud solution vendors prepackage a set of rules or model labeled for each channel, but there is little or no sharing of the details of individual transactions across channels. Only a model built upon the history of integrated, detailed, cross-channel data will leverage cross-silo benefits. Such models are necessarily custom because of the variations in multi-channel data among payment operators.
If you’re being sold licenses for each channel, it’s not truly cross-channel
Vendors selling licenses-by-the-payment-channel, really don’t have a cross-channel solution but rather a “multi-channel” approach. The data model needs to be flexible and configurable to adapt to multi-channel data variations. So, any vendor offering a fixed data model is not truly accommodating a cross-channel solution. Moreover, vendors selling solutions by the channel have clearly componentized their solutions where pre-packaged predictive models and data models are segmented, channel focused, and don’t provide a real cross-channel solution.
This segmentation is basically a multi-channel, “division of labor” approach where data sets are independently monitored and assessed and are only later combined and processed in unison. Similar to spiders, with several sets of eyes, each set for a specific purpose and detection of movement, contributing independently to painting the full picture of the surroundings. But even with the quantitative advantage, spiders, in general, tend to have poor overall eyesight. The human eye, in contrast, captures all available data and movement in one instance and processes the full picture.
Similarly, IBM Safer Payments’ data integration is channel independent, fully able to combine detailed data from multiple channels across time and transactions, and IBM Safer Payments has the built-in model-building capability to make sense of fully integrated cross-channel data and, based on machine-learning, determine the riskiness of cross-channel patterns. IBM Safer Payments does not offer an ill-ﬁtting standard solution for each of several payment channels. It is designed to integrate data across channels and build models using that integrated data to implement true enterprise fraud management. For more information on our solution, visit www.ibm.com/saferpayments