Secure Banking Transactions Through Real-Time Fraud Detection Using Apache Kafka and Machine Learning

Abstract

The rapid growth of digital banking has increased the number of online transactions, making fraud detection an important challenge for financial institutions. This project presents a real-time bank transaction fraud detection system that combines Apache Kafka with Machine Learning to identify suspicious transactions as they occur. Apache Kafka is used to stream transaction data continuously, while a Random Forest classifier analyzes each transaction and predicts whether it is genuine or fraudulent. Before training the model, the transaction dataset is cleaned, encoded, and normalized to improve prediction accuracy. A Flask-based web application provides an easy-touse interface for generating transaction streams and displaying fraud detection results in real time. The proposed system reduces the delay associated with traditional batch-processing methods and enables faster decision-making. Its modular design also allows the system to handle increasing transaction volumes, making it suitable for modern banking environments that require secure, reliable, and scalable fraud detection solutions.