Feasibility study and implementation of paper “fingerprint” anti-counterfeiting system using flatbed scanner

Abstract

Counterfeiting is a tough issue with serious negative impact to the society, that people are always trying to resolve over the years. Presently, most anti-counterfeiting techniques have its own limitations. The idea of using paper’s natural unique three-dimensional surface texture for authentication shows a new perspective. Among the different researches done on this topic, Clarkson et al. (2009) proposed a novel idea of using the commodity scanner to scan the paper at four orientations and processing the scanned images to obtain the normal information for each surface point, so as to use it for feature vector computation. This Final Year Project is mainly inspired by the idea and framework proposed by Clarkson et al. in 2009 with proper modifications and improvements. A feasibility study on this topic was done to see whether the unique paper texture information can be captured with even cheaper devices and used to distinguish different papers, as cost and workability are two important factors affecting whether the system can be popularized. With the result showing its feasibility, the actual general-purpose paper “fingerprint” anti-counterfeiting system prototype was developed with the goals of economic, accurate, robust, secure and widely applicable. Several improvements have been made. Firstly, the system can accept images scanned using extremely cheap flatbed scanner with any scan resolution larger than 75dpi and smaller than 1200dpi and saved in different formats. Secondly, it does not require any modification or marking on the original paper for alignment. At the same time, many important findings that need special attention during actual implementation are pointed out: The existence of different error sources were discovered, which are to be carefully handled during implementation; The need to align the validation images with the image I0˚r (I0˚r refers to the I0˚ image used during registration) is highlighted, and the solution on how to align them without the need of storing I0˚r is suggested; The potential issues that users might not know which side of the paper should be scanned for validation and which image should be labeled as I0˚v (I0˚v refers to the I0˚ image used during validation) are carefully considered. The paper “fingerprint” anti-counterfeiting system prototype was developed with the goals in mind, and it is proven that the goals are met successfully after doing the robustness test. In addition, some of the possible applications of the system are explored, which again shows its value. Nevertheless, there are still space for further improvements. Possible future works are listed in this report showing the system’s potential to reach a higher level.