Machine Learning Is New Weapon Against Bank Fraud

AI tools that detect attempted fraud and collect evidence to convict and deter criminals are becoming a key strategy—and a large expenditure—for banks. In recent years, the risk of being victimized by fraud makes banks vulnerable to large fines and damage to their brand.

Several trends have made fraud prevention and detection an urgent issue for banks. First, there are more customer channels for bad actors to attack, including mobile payments. Next, criminals have become more adept at using malware and phishing emails to penetrate security and access customer accounts. Finally, legacy technology can’t identify and respond fast enough to new kinds of fraudulent activity. Outdated platforms and tools force bank personnel to spend too much time investigating “false positives,” transactions that appear suspicious but are in fact legitimate.

The challenge for banks: correlate and analyse a mountain of data from multiple systems to identify anomalous activities that could represent fraud—and then respond swiftly.

Aided by increasing processing power, Big Data and advances in statistical modeling, Machine Learning (ML) has become an important tool to meet these challenges. ML algorithms, also called neural networks, are basically advanced analytics that run continuously in real time against transactions and related data like speech, images, text and online behavior. The algorithms identify patterns that suggest controls may have been compromised. They flag questionable transactions for immediate review by investigators.

ML algorithms are more efficient than older, rule-based anti-fraud models. They are adaptive, able to learn and adapt to new fraud patterns by continuously testing themselves against legitimate behavior. ML-based methods reduce the volume of false positives, leaving human experts free to be deployed on higher-value detection and prevention.

The scope of anti-fraud activity needs to expand constantly. Case in point: as voice-based bank services become more important, new tools are required. In 2017, the European Contact Centre & Customer Service Awards (ECCCSA) gave its prestigious prize to Lloyds Banking Group and Pindrop, a developer of phone fraud prevention technology. Pindrop analyzes call features such as the location.