AI-Powered Risk Scoring and Fraud Detection Frameworks in Real-Time Payment Processing Systems

Abstract

The increasing adoption of electronic payments has given rise to new opportunities for fraudsters, posing significant challenges for the financial service industry. In this context, innovative countermeasures are essential to address their criminal activities. Card-not-present (CNP) payments, such as Internet payments or payments made over the phone, are characterized by a higher percentage of fraud attempts and a higher average number of fraud attempts per person [1]. Consequently, automatic systems are needed to analyze previous transactions and detect those transactions that deviate from the standard behavior of the user within a limited time. These systems are often designed to identify fraud by taking into account fraud behavior in a global view. However, schemes that prohibit certain numbers of transactions in a specific time window and analyze transactions directly without any consideration of the user account performance have been proposed.