The rapid growth of digital payment systems has significantly increased the risk of financial fraud, making traditional
rule-based detection methods ineffective against modern, intelligent attack patterns. Existing machine learning approaches
improve detection accuracy but often lack transparency, auditability, and trust, especially in regulated financial environments.
To address these challenges, this work presents an integrated framework that combines Artificial Intelligence, Explainable AI,
and Blockchain to deliver a secure, transparent, and real-time fraud detection system.
The proposed framework analyzes each transaction using an AI-based risk scoring model capable of identifying rare fraudulent
activities in highly imbalanced datasets. Along with prediction, an explainability layer generates human-understandable reason
codes for every decision, improving interpretability for users and auditors. To ensure data integrity, all critical fraud evidence—
such as transaction details, model outputs, and explanation vectors—is stored off-chain and anchored on a blockchain using
cryptographic hashing, making it tamper-proof and verifiable.
The system also introduces a structured dispute resolution workflow, enabling transparent tracking of cases from initiation to
resolution. Additionally, the framework is designed to operate in real-time, maintaining an end-to-end processing latency of
approximately 0.25 seconds per transaction. By combining high-recall detection, explainable decision-making, and immutable
audit trails, the framework provides a comprehensive solution for enhancing trust, compliance, and security in modern digital
payment ecosystems.