Fraud detection is one of the most pressing issues in banking. The U.S. Federal Trade Commission has logged millions of fraud-related complaints in the last year so it is evident that the scale of the problem is incredible. As such, the potential gains from successful prevention are equally massive.
The Problem
It is very difficult to build an accurate fraud system that works in real-time, as the number of fraudulent card charges is so much lower than the total volume (VISA alone processes over 2,000 transactions every second) that establishing a training set is difficult.
Traditional rules based systems are insufficient, often resulting in false positive rates that exceed 90%. This creates a massive number of false positive alerts which then need to be cleared through human intervention. Repetitive analyst action can lead to a "numbness to false positives," raising operational risk and subsequently regulatory risk in the process.
An Obstacle
So long as there are banks, there will be those who try to defraud them. However, anomaly detection offers sophisticated techniques to combat fraud. Anomaly detection is the process of finding patterns of interest (outliers, exceptions, peculiarities, etc.) that deviate from expected behavior within datasets. 
However, anomaly detection models require vast amounts of relevant data in order to function. This means that data pipelines must process information in real-time, no easy feat. Additionally, data needs to be relevant to the population of users and transactions in question; data from Singapore cannot much help combat fraud in Mexico City.
Since financial data has highly rigorous compliance regulations, building a financial model on vast, relevant data without siloing teams faces added complications. Yet it is necessary in order to protect user (and stakeholder) interests.
The Solution
As an application domain within anomaly detection, machine learning-based fraud detection use cases dominate the banking industry. ML-based fraud detection uses anomaly detection to uncover behavior intended to mislead or misrepresent.
When well-integrated into operations and supplied with real-time data, anomaly detection can support a safe user experience. Banking organizations will thus be able to secure users’ trust and drive secure transactions.