Real-Time Fraud Detection in Digital Transactions Using AI-Powered Anomaly Detection Models

Abstract

The exponential growth of cryptocurrency transactions in digital economies has increased the risk of fraudulent activities that threaten transparency and trust. This paper proposes a real-time fraud detection system that integrates artificial intelligence (AI) and blockchain-based cryptographic hashing for enhanced transaction security.

The main aim of this research is to detect suspicious patterns dynamically while ensuring data integrity. The study’s objectives include analyzing user behavior, identifying anomalies in digital transactions, and minimizing false alerts through ensemble learning. The system utilizes 100,000 real-world cryptocurrency transaction records, combining Support Vector Machines (SVM), Decision Tree (DT), and Neural Network (NN) models with a logistic regression meta-learner.

Experimental results demonstrate that the ensemble model achieved an accuracy of 96.8% and reduced false positives to 3.9%, outperforming single models. The study concludes that combining AI and blockchain provides a secure, adaptive, and scalable solution for real-time fraud detection, contributing both theoretical insights and practical applications for the fintech ecosystem.