AI-Driven Fraud Detection: A Risk Scoring Model for Enhanced Security in Banking
Shubham Metha *
M.S. in Information Systems Northeastern University, United States.
*Author to whom correspondence should be addressed.
Abstract
As technology makes advancements so does the risk of accessing it for wrong doings. In recent years as we moved from traditional banking systems to online banking and the volume of digital transactions has increased eccentric. This also comes up with increasing risk of fraudulent activities like accessing bank accounts, credit card frauds, account frauds, dormant account fraud, and many others. Detecting fraud activities is and crucial part of banking system.
This research explores the application of artificial intelligence (AI) in detecting potentially fraudulent activities by generating a risk score to assess account behavior. A formula is developed to compute a score out of 100, which triggers automated security measures when exceeding a predefined threshold of 80. The formula evaluates four key activities commonly associated with fraud: new device logins, updates to contact number or email address, the addition of new payees or Zelle contacts, and transactions exceeding $1,000 in 48-hour time span.
Leveraging machine learning algorithms, this model incorporates behavioral patterns, historical data, and real-time anomaly detection to calculate the score. Accounts with scores above the threshold are temporarily locked, initiating further verification processes to ensure security while minimizing customer inconvenience. This research demonstrates the effectiveness of AI-driven fraud detection mechanisms and highlights the balance between security and user experience in modern banking systems.
Keywords: Artificial intelligence, machine learning, customer protection, banking, security, fraud, risk, customer satisfaction, business growth, finance, online banking, account takeover