Detection of Deep-fake QR Codes in Street Vendor Digital Payment Systems Using Frequency Domain Analysis

Abstract

QR code-based digital payments are widely used in street vendor transactions, especially in India through UPI systems. These payments are fast and easy, but they also face security risks. One major problem is the use of fake or deep-fake QR codes. Fraudsters replace or modify original QR codes and redirect the payment to their own bank accounts. Customers and vendors often do not notice this change, which leads to financial loss. Most existing research focuses on detecting fraud after the transaction is completed using machine learning models. Very few studies focus on detecting manipulated QR codes before the payment happens. This paper proposes a new method to detect deep-fake QR codes using frequency domain analysis. Instead of only checking the visible structure of the QR code, the proposed method studies hidden frequency patterns using techniques like “Fast Fourier Transform” (FFT) and “Discrete Cosine Transform” (DCT). When a QR code is edited or tampered with, small changes occur in its frequency components. These changes are not clearly visible to the human eye but can be detected through frequency analysis. The proposed system is lightweight and suitable for real-time use in street vendor payment systems. Experimental results show that the method can effectively identify manipulated QR codes with good accuracy and low computational cost. This approach can improve security in QR-based digital payments and help prevent fraud in small-scale business environments.