AI-Based QR Code Fraud and Scam Detection System for Secure Digital Payments
AI-Based QR Code Fraud and Scam Detection System for Secure Digital Payments
Muralidharan V 1, Saran R 2 , Vignesh Perumal L S 3 , Agesta Jenifer A4
1 PG-MBA & Karpagam College of Engineering – SoMS, Coimbatore, India
2 PG-MBA & Karpagam College of Engineering – SoMS, Coimbatore, India
3 PG-MBA & Karpagam College of Engineering – SoMS, Coimbatore, India
4 UG-CSE & Holy Cross Engineering College, Thoothukudi , India
---------------------------------------------------------------------***--------------------------------------------------------------------
ABSTRACT - The rapid proliferation of QR code-based digital payment systems in India, accelerated by the Unified Payments Interface (UPI) ecosystem, has been accompanied by a sharp rise in QR code-based fraud schemes. The Reserve Bank of India's Annual Report on Digital Payment Frauds documented that UPI-related fraud cases increased by 85% between fiscal years 2022 and 2024, with QR code manipulation, fake merchant codes, and quishing (QR code phishing) attacks emerging as dominant attack vectors, particularly targeting small merchants, elderly users, and first-time digital payment adopters with limited technical literacy. Conventional QR code scanning applications perform no security validation beyond basic format parsing, leaving users exposed to malicious payloads embedded within seemingly legitimate-looking codes. This paper presents an AI-Based QR Code Fraud and Scam Detection System that performs multi-layered real-time security analysis at the point of QR code scanning, combining a Convolutional Neural Network-based visual tamper detection module that identifies physically altered or overlaid QR codes, a URL and payment payload risk classifier using a fine-tuned BERT-based natural language model trained on a labelled corpus of fraudulent and legitimate UPI payment strings, a merchant verification layer that cross-references VPA (Virtual Payment Address) and merchant identifiers against the National Payments Corporation of India's verified merchant registry, and a behavioral anomaly detection module that flags unusual payment patterns such as sudden high-value requests from previously low-value merchant codes. Evaluated against a dataset of 38,400 QR codes comprising verified legitimate merchant codes and documented fraud samples collected in collaboration with cybercrime cell reports from three Tamil Nadu districts, the proposed system achieves a fraud detection accuracy of 96.7%, false positive rate of 2.3%, and end-to-end scan-to-verdict latency of 340 milliseconds, demonstrating strong viability for integration into existing UPI-enabled payment applications.
Keywords — QR Code Security, UPI Fraud Detection, Quishing, BERT, CNN, Digital Payment Security, Phishing Detection, Merchant Verification, Behavioral Anomaly Detection, Fintech Security