How Retail Banks Can Fight Account Opening Fraud with AI

Banks are seeking new ways to improve revenue, reduce risk, and create a stronger customer experience. But these efforts open up new attack vectors for today’s determine fraudsters, who are evolving faster than the detection strategies designed to stop them. Many banks are finding that their traditional approaches to fraud management are falling short—and leaving them exposed to risk.

The limitations of rules

Fraud moves fast. Fraudsters are opportunistic, seeking out a security system’s biggest vulnerabilities and leveraging weak points through coordinated attacks. Banks have traditionally relied upon rules-based fraud detection systems to counter these threats, but fraud advancements have outpaced the capabilities of these systems.

Rules-based systems tend to be either too broad or too narrow in scope to adequately address fraud attack vectors, requiring banks to combine multiple solutions into a single system to cover their bases. These patchwork fraud detection systems address only specific pain points in the bank’s system and fail to fully protect the institution from new threats that may emerge.

Customer experience is also debilitated by these combination solutions (e.g. combining biometric authentications such as device fingerprinting with binary checks such as email validation). Each of these systems has its own set of rules, and as new solutions are added, these rules multiply to the point where a unified, seamless customer experience is impossible to achieve.

Augmenting rules with machine learning

Given that rules-based systems are based only on the historical data available to the enterprise, they’re ill-suited to protect banks against the multi-faceted fraud attacks we see today. Account opening fraud is the best example: How can retail banks go toe-to-toe with fraudsters without the necessary historical data required for effective rules-based systems?

The answer is machine learning. In many ways, machine learning is the natural next step beyond rules-based systems, combining historical data and advanced algorithmic assessments to make decisions about current—and future—behaviors.

Machine learning doesn’t replace rules completely, but it complements them to expand the capabilities of the risk management platform. And when applied to large datasets like those found in account opening analyses, these algorithms can pinpoint surprising and unintuitive fraud signals. For example, Feedzai’s machine learning has found that:

  • Devices with high battery power are correlated with higher rates of fraud;
  • Email addresses with two to four consecutive digits are more likely to be fraudulent by three to four percent;
  • Specific email domains feature higher correlation with fraudulent behavior, including public domains;
  • Devices with unknown or null names are fraudulent 78 percent of the time.

To the human eye, these insights are merely bits of trivia that are impossible to put into action. But correlated together in real time by a machine, they’re data points that provide valuable information on fraud risk for every new account opened.

A holistic approach to risk management

The only way to get ahead of account opening fraud is to take a holistic approach to risk management that’s enabled by AI. These advancements are no longer optional; without a machine learning-based system that combines data from internal and external sources, financial institutions won’t be able to keep up with the fast-moving fraud landscape. Banks need to be aggressive against financial crime, and adopting machine learning is the best way to start.

stefan jandreau-smith

Product MarketingFeedzai

Latest posts by stefan jandreau-smith (see all)