Unmasking Fraud: Applying Machine Learning to Detect Bank Drops

This paper evaluates eight machine learning (ML) models that are capable of flagging bank accounts as fraudulent at the time of bank account registration.

Bank drops are fraudulent bank accounts that are used to funnel and transfer illegal funds. Most fraudulent bank accounts often go undetected, exposing flaws in the banking system and require large teams of cybersecurity specialists to spend significant amounts of time and money trying to correct. This paper evaluates eight machine learning (ML) models that are capable of flagging bank accounts as fraudulent at the time of bank account registration. Each of these models process bank account registration information to instantly determine the legitimate versus fraudulent status of an account, expediting the bank account flagging process, saving banks internal resources and protecting bank clients from fraudulent activities. The highest performing models are ensemble models such as the gradient boosting classifier and the random forest classifier, each having accuracy scores above 90%. These results show that ML models are a valuable tool to detect fraudulent accounts and protect banks across the world, effectively and efficiently stopping crime before it begins.

Explore More!

Research Paper