Predicting Human Trafficking – The Power of Analytics
May 20th, 2016
Not too long ago, during a flight back home from a customer site, I overheard a flight attendant talking to her colleague about recent training they received on the topic of human trafficking. Flight attendants are in a position to detect suspicious passenger behavior associated with human trafficking such as young passengers escorted by adults who don’t behave like family members; passengers going on long haul flight empty handed without even carrying small personal items; passengers escorted by companions that speak on their behalf or who refuse to leave them alone (even to go to the restrooms); and passengers who express unusual subordinate and fearful behavior in comparison to other passengers. When flight attendants suspect that any of their passengers may be the victim of human trafficking, they often call ahead to the authorities to intercede and prevent a tragedy from happening.
In the financial services world, things are more complicated because the type of financial behavior demonstrated by human traffickers is often similar to legitimate business activity. Still, there are early warning signs that can often help detect the presence of trafficking.
Let’s take, for example, the process of when a human trafficking ring is establishing itself in a new region. The type of activities we would be looking for could include the bulk opening of bank accounts in several branches that are close to borders. We could also see an increase in purchases of pre-paid cards, as these are often used to pay for lodging, food and smugglers’ expenses. These activities, combined with international wires being sent to high risk geographies with known large high migrant population, may indicate that we’re looking at a potential human trafficking network in the process of being setup.
The problem with confidently detecting these activities is that each transaction by itself might look like a legitimate transaction, and that the volumes in this initial setup may not be significant enough for rule-based systems to generate alerts.
This is where we can leverage the power of advanced analytical methods. With the usage of network-based analytics, sophisticated algorithms can be used to identify a multiplex of entities and their various linkages. This allows us to identify the burst in these activities not just as isolated events, but in the context of the rest of the activities within the network.
Predictive models can be trickier to apply in this case since they require a significant amount of data about existing cases — for example, at least several hundred cases that resulted in proven activity of human trafficking in the past might be necessary to in order to “teach” the model this is an abnormal activity. But even without massive amounts of data, predictive analytics techniques can be used effectively to better identify peer group allocations, elevating scores given for abnormal activity used in the trafficking organizations with the usage of anomaly detection models.
With the early warning of these activities that predictive analytics can provide, financial institutions can put newly established accounts under a higher level of scrutiny, monitoring for patterns that we might see appear in later stages such as high volumes of money movement to high risk geographies perhaps combined with high amount of pre-paid cards activity to make sure the activity is detected and reported.
But we can do one more thing with the early warning which is to alert the relevant authorities – even at the early stages of the ring being setup, with the hopes that this will be able to prevent additional activity of human trafficking and reduce human tragedies.