Fraud analytics professional
Payment card fraud detection involves banks and financial organizations leveraging analytics to delve into card transaction data and build counter-fraud models. Over the years, the discipline of fraud analytics has evolved to crafting machine learning algorithms which recognize highly complex fraud patterns in enormously large volumes of records.
Meanwhile, the variety of data sources being used to detect fraud has grown from just card use details to information related to customer internet banking, social media activity, device profiles and more.
Of late, smart cameras, CCTV devices and drones with intelligent cameras are amassing vast expanses of video data at extraordinary scale and speed. The immense volume of video produced renders it impractical for effective manual processing. Hence, increasingly advanced Deep Learning algorithms are being developed to automatically process and analyze these vast streams of footage.
Against the backdrop of such technology shifts, video analytics is rapidly emerging as a powerful tool to fight fraud. To illustrate, ATM fraud can be mitigated by writing software codes that routinely identify irregular patterns in vestibule video feed. Hypothetically, a crook using a stolen card to withdraw cash might display cues in body language quite distinct from that of a genuine card holder, such as being nervously aware of the ATM camera, attempting to cover their face or other similar behaviors. Machine learning algorithms working on such video inputs could quickly recognize suspicious patterns in body language and raise alarms automatically. Metadata from video files, such as time spent in cash withdrawal, can also be leveraged. Facial recognition techniques when applied to video data has the potential of identifying repeat offenders, by comparing against a database of previously identified criminals.
In addition to ATM fraud detection, video analytics can empower merchants by employing in-store camera feeds to routinely spot irregular patterns in floor movements and body language of card users. To take this idea even further, video analytics also has the potential to uncover possible relationships between cashiers and fraudsters. Alerts generated from video inputs could then be used in conjunction with those from traditional data sources, such as card transactions, to arrive at a holistic decision about any transaction being fraudulent or not.
Video analytics can also help glean valuable insight about internal fraud by automatically identifying suspicious aspects in employee behavior such as entering or exiting the office at unusual times or attempting to access restricted areas.
As fraud examiners on the front lines, we know fraudulent activity contributes to reputational risk, in addition to the prospect of monetary loss. In a world where criminal schemes continue to rapidly evolve, it is increasingly pertinent to build a fraud-fighting infrastructure, while also leveraging unique and emerging sources of data.
Chatterjee is a fraud analytics professional at a global bank and has spent more than a decade designing counter-fraud solutions.