An Application of Deep Learning Technology in The Recognition of Forged Documents with Color Laser Printing

An Application of Deep Learning Technology in The Recognition of Forged Documents with Color Laser Printing

Advanced laser printing technology is now widely used in daily life and at work. Because laser printers are so easy to obtain and use, they have also become a popular method for criminals to forge important documents like currency, securities, and certificates. To obtain evidence for document forgery, traditional forensics use chemical analysis of ink, but the cost is high. Laser color printing uses halftone technology. This technology combines the CMYK printing colors’ halftone screens at their respective angles. When correctly superimposed, they will form a moiré pattern. And the moiré pattern of a color block differs depending on the printer brand or model. This paper will use deep learning to create an image recognition model that can quickly classify forged documents by combining image processing, data augmentation, and transfer learning. The experimental results are examined using indicators of accuracy, precision, recall, F1-score, and PR curve. The results show that ResNet-50 has the best performance and can be used to detect document forgery. The study’s findings will be useful for future criminal investigators and forensic personnel in their work.