Machine Learning’s Role in Protecting Mobile Payment Transactions
Originally posted on the Mobile Payments Conference blog.
Mobile payments are synonymous with today’s digitally-connected world. At the forefront of mobile payment innovation is the introduction and branding of company-specific mobile apps. Mobile apps provide a gateway for organizations to engage with customers in a digital environment, monetize the interaction and build brand loyalty with enhanced ease of use and personalization. On top of all these benefits, today’s mobile apps are digital sales platforms that are catalysts to the continued growth of the mobile payments industry.
With all the promise that mobile apps and payments present, there also comes risk. Specifically, merchants today need to account for increased attacks and fraudulent activity. Industry statistics presented by Kount partner Cayan, recognized that CNP fraud in the U.S. is expected to hit $6.4 billion this year in comparison to $3.1 billion in 2015.
This staggering leap is only expected to escalate as merchants migrate online and criminals continue to follow the money. Three common forms of CNP fraud include:
- Account takeover (ATO): ATO is growing rapidly in the industry because of its low cost of entry and minimal risk of getting caught. Specifically, ATO fraud is used to fraudulently login to an existing account. Once established, a fraudster will take advantage of the value of that account. This may include using the saved payment method or loyalty points to make purchases.
- Loyalty reward points: Because reward points work like cash, fraudsters will try to find weaknesses in the system and steal reward points to sell. This includes account takeover and brute-force attacks. Once the fraudster gains access, the subsequent fraud can be difficult to detect.
- eGift cards: Electronic gift cards are easily converted into cash, a key requirement for fraudsters. Considered “low hanging fruit” for fraudsters, eGift cards can represent thousands of dollars and be sold at a discount to easily convert them into cash.
To address these three schemes and many more, merchants need to deploy a fraud strategy that is multi-layered and specifically accounts for card-not-present fraud. By addressing fraud with a holistic strategy, merchants can authenticate the user, identify fraudulent behavior and stop fraud before it influences the bottom line and diminishes the merchant’s brand.
An underpinning technology to stopping CNP fraud is machine learning. Machine learning combines data, context, and feature engineering to allow organizations to authenticate the person who is behind the submit button. Machine learning, a form of artificial intelligence, allows fraud prevention solutions to “learn” on their own and continually improve results without human intervention. For stopping card not present payment, there are two critical types of machine learning that, when combined, provide the best fraud prevention foundation.
- Unsupervised Machine Learning. Unsupervised learning does not require outcomes, so it can learn without waiting for the three-month chargeback reporting cycle. This type of learning relies on clustering, peer group analysis, breakpoint analysis or a combination. This enables fraud prevention solutions to detect patterns and anomalies within extremely large sets of data.
- Supervised Machine Learning. Supervised learning uses outcome-labeled training data sets to learn. Models include neural networks, Bayesian classifiers, regression, decision trees, or an ensemble combination. Massive amounts of data run through defined models to assess risk outcomes.
The use of machine learning within a fraud solution empowers decision makers the ability to access data, understand its meaning and make informed decisions to stop fraud before it impacts the businesses’ bottom line and the overall brand. The advanced analytics performed on the collected data in relation to industry data provides the merchant with the ability to decide in a real time manner the authenticity of the individual.
As machine learning continues to operate behind the scenes to protect merchant and customer transactions, it is critical that organizations stay abreast of evolving fraud tactics and trends. Today’s merchants must monitor the trends of fraud within their own vertical market and others, tap the resources of fraud experts and think holistically as it relates to implementing fraud prevention strategies.
To learn more about Kount's patented and proprietary machine learning technology, download the eBook "The Truth About Machine Learning in Fraud Prevention".