Digital Forensics of Scanned QR Code Images for Printer Source Identification Using Bottleneck Residual Block

Area of Science:

• Digital Forensics
• Computer Vision
• Machine Learning

Background:

• QR codes are ubiquitous, impacting daily life and commerce.
• The vulnerability of QR codes to printing and forgery poses risks of economic loss and criminal activity.
• Accurate identification of QR code printer sources is essential for security and authenticity verification.

Purpose of the Study:

• To propose a novel method for identifying the printer source of scanned QR code image blocks.
• To enhance the security and trustworthiness of QR code applications.
• To develop a robust deep learning model for digital image forensics.

Main Methods:

• A convolutional neural network (CNN) based method named PSINet (printer source identification network) was developed.
• PSINet incorporates a bottleneck residual block (BRB) for improved feature extraction.
• The study provides theoretical discussion and experimental analysis of PSINet's architecture and input design.

Main Results:

• PSINet achieved exceptional performance in printer source identification for QR codes.
• The method reached an accuracy of 99.82% when tested on eight different printers.
• PSINet outperformed established models like LeNet and AlexNet, as well as other state-of-the-art deep learning approaches.

Conclusions:

• The proposed PSINet demonstrates superior effectiveness for QR code printer source identification.
• This method offers a significant advancement in digital image forensics and QR code security.
• PSINet provides a reliable solution to combat QR code forgery and related criminal activities.