A Deep Learning Approach for QR Code Based Printed Source Identification

With the widespread popularity and its ease of use, QR codes are relatively easy to be reproduced or forged illegally through printed documents. One solution to tackle this problem is to identify the source printer which produced the QR codes. In this paper, we study the CNN models for QR code based printed source identification through a series of experiments which involved with grayscale QR codes. Our experimental results show that the pretrained CNN models such as AlexNet, GoogleNet and ResNet could identify printer source with high accuracy even though the models are trained with limited input dataset.