Are elder abusers the ultimate insiders?

Teradata Guided Analytics
Teradata Employee

Earlier this month, a colleague forwarded a report that the Consumer Financial Protection Bureau released earlier this year. The upshot is that the CFPB wants retail banks and credit unions to pay more attention to elderly customers to help reduce the problem of elder financial abuse.

 

The elderly are attractive targets since they have accumulated assets over time and may be facing declining ability to detect and prevent abuses, especially from those entrusted to assist them. Recommendations from the CFPB include increased training of tellers to take note of suspicious activity and to report anything to the Adult Protective Services. This is acknowledged to be a just a first step in this effort.

 

Most relevant for data scientists is the prominent recommendation to use technology to monitor for signs of elder financial exploitation.

 

Because indicators of elder fraud risk may differ from conventionally accepted patterns of suspicious activity, financial institutions using predictive analytics should review their filtering criteria against individual account holders’ patterns and explore additional risk factors that may be associated with elder financial exploitation"

 

This closely mirrors the approach we use in our Insider Threat solution. In the case of Insider Threats we are focused on the corporate network and the information assets housed in various data repositories. A person’s network access, and data movements (among other things) are inputs into our behavioral analytics to generate a risk rating for any particular person or activity. Business users provide the rules and business logic particular to their particular goals to determine the thresholds for the predictive model’s results.

 

It’s not difficult to adapt this work to focus on elder financial abuse. Historical account activity can be used in aggregate to establish a baseline for comparison. Account access, profile changes, and characteristics of asset movements can serve as inputs to the analytic process to help predict whether any particular transaction should be cause for alarm.

 

A critical challenge of the Insider Threat case that also applies here is that activities and behaviors, in isolation, do not generally set off alarms. The activities in both cases are usually authorized and allowed by existing safeguard systems. Caregivers/family members are often put in charge and are authorized to handle financial matters on behalf of the elders being abused. We face this challenge by utilizing both depth and breadth in the available data and by incorporating these into the risk calculations. The actor’s profile, geographic location, timing, and other characteristics (such as fund destination) of each and every transaction can be aggregated, scrutinized, and compared to reveal anomalous account activity that warrants attention. This big picture approach is necessary not only to fine tune the predictive capabilities of the model and keep the error rates in acceptable ranges, but also to allow us to train and improve the model sophistication over time as true and false predictions get incorporated into future iterations.