Araliya Sammé, Head of Financial Crime at Featurespace, writes for Banking Strategies about how Joe Exotic’s infamy exposes the pervasiveness of much larger problems facing the financial services industry (read the full article here).

The popular documentary “Tiger King” on Netflix gripped the world with its in-depth look at the business of exotic animal trafficking and the character of those involved. This billion-dollar business is often linked with other types of organized crime, such as terrorism, drug trafficking and arms trading. Around the world, criminals exploit wildlife to generate funding, with support provided by cash produced by other crimes. This is a part of the perpetual ecosystem that is money laundering.

Banks are bound by law to highlight suspicious customer activity, so it is imperative to be able to identify, evaluate and track criminal behavior so that the flow of financials can be traced to the source. Criminals, especially those within sophisticated groups, change their tactics regularly to avoid detection. Collaboration between the crime fighters looking out for this illegal activity allows for shared intelligence to catch criminals, who may move from bank to bank to evade detection. With shared information, financial crimes investigators would be aware of suspicious behaviors that are indicative of specific crimes – such as animal trafficking – and be able to intervene much earlier.

Machine learning and adaptive behavioral analytics, which can monitor and recognize anomalies across the enterprise, can be vital tools in this effort.

A challenge is that advanced machine learning models can only iterate through the different indicators if high-quality data is being fed through, and most financial crime experts admit that such data difficult to obtain. The adage “garbage in, garbage out” rings especially true in this scenario.

Spotting specific crimes like human and wildlife trafficking, drug smuggling and terrorist financing is notoriously difficult. It requires experts who know the clues – such as information on cross-border movement of money, animals or people, shipping and customs documents and networks of entities – that indicate complex criminal networks. These clues can vary from group to group and location to location.

Working in conjunction with experts who have specific knowledge of certain types of financial crimes ensures an appropriate machine learning model is built, that the right data is being fed into it, and that the proper logic is applied in real time to thousands of transactions per second.

The model is built to monitor for known suspicious behaviors, such as transacting with a high-risk jurisdiction or an unusual purchase. In these cases, the behaviors may be legitimate – anyone can make an impulsive purchase — so the model must accurately identify the risk without flagging every bit of activity as a false positive that analysts must investigate.

As behaviors change, the models adapt and assess the legitimacy of new activity based on recent data and, while this happens automatically, an analyst can also step in and adjust a model to improve its performance.

Continue reading this article in Banking Strategies.