Printer source identification by feature modeling in the total variable printer space

Abstract

Although office systems are moving toward paperless electronic structures, paper documents are still of high importance in everyday communications [1]. Paper documents are used as direct tools in identity identification, taxes, and financial transactions. Sources of these documents are generally laser or other types of printers. Public access to these printers and developing image editing software have facilitated editing and reproducing similar digital printed documents [2]. Forgery and tampering of paper documents may destroy general trust; therefore, it is very important to recognize forged printed documents and ensure their authenticity. Printer source identification is one of the forensic methods for ensuring the validity of scanned paper documents [3]. Authenticity of a printed document can be identified by a forensic expert and chemical tests; however, this process may destroy the printed document [4]. Therefore,

Document forgery detection using source printer identification: A comparative study of text‐dependent versus text‐independent analysis

“…Nevertheless, it can be observed that a major proportion of studies have been evaluated in a textdependent framework (Ferreira et al, 2015;Shang et al, 2014;Tsai et al, 2015;, and in many cases, on isolated characters. Though identification rates close to 100% have been reported (Hamzehyan et al, 2021;Tsai et al, 2019), the numbers are not directly comparable with those reported by our system due to a different number of printers and a different experimental setup. Furthermore, the results of text-dependent and text-independent analysis cannot be directly compared.…”

“…Classification using the K‐nearest neighbour classifier reported accuracy values of 87% and 91% with full and reduced feature sets, respectively. In another recent work based on textural features (Hamzehyan et al, 2021), a document image is divided into patches and primary features based on local binary patterns are extracted from each patch. Secondary features are then extracted using joint factor analysis, allowing each document to be represented in a low‐dimensional feature space.…”

“…The recent past has witnessed a digital revolution where many businesses and processes are moving towards digital in lieu of printed documents. Despite these developments, printed documents remain an integral part of many organizations (Hamzehyan et al, 2021). This is primarily attributed to the easy accessibility of a wide variety of low‐cost printing devices.…”

“…In some cases, printed and digital documents co‐exist due to ease of use and security challenges (Joshi et al, 2022). Nonetheless, with the advent of easily accessible printing and scanning facilities and state‐of‐the‐art image manipulation tools, an increased frequency of alteration of printed documents for malicious purposes has also been observed (Hamzehyan et al, 2021; Joshi & Khanna, 2017). Typically, printed documents are first digitized and then forged using image‐editing software (Kumar et al, 2020).…”

Authentication of rotogravure print-outs using a regular test pattern

“…By analogy to image forensics, this approach aims at identifying the printer and possibly the scanner [2] that were used to produce a given hardcopy document/packaging and its scanned version. The printer and captured device identifications are more complex problems than camera identification as some mechanical and optical components are involved in the process of captured document production chain [11]. Some recurrent features are extracted from printed characters according to different techniques as gray-level co-occurrence matrix [23], [24], noise energy, contour roughness and average gradient of characters edges [34].…”