Printed document forgery detection using text reordering and mixing of matrices in zero watermarking

Abstract

Printed documents are still needed in our daily life even though we are living in digital era. Information either in printed document or in digital form must be protected from threats and attacks such as forgery or unauthorized modification. Such threats makes the document lose its integrity and authenticity as well as the value of the information. There are several methods that have been used and created to maintain authenticity, integrity and detecting forgery of printed documents. However some of the methods are not suitable for public use due to its complexity, hard to obtain special materials to secure the document and expensive. This project studies on several text watermarking methods that have been used for document security. There are four main approaches in text watermarking with their own advantages and disadvantages. Zero watermarking is another simpler yet effective method to verify the integrity and to detect forgery of text document which can be used as an alternative to text watermarking. Based on the studies that have been conducted, a zero watermarking algorithm was proposed to improve weakness found in one of the content-based zero watermarking algorithm. This document forgery detection solution uses text content to generate watermark by implementing text reordering and mixing of matrices. The generated watermark is registered into the trusted third party instead of embedding the watermark into the text document. The generated original watermark stored in the trusted third party is compared with extracted watermark from the received text document later on for verification and forgery detection. A performance analysis of the proposed zero watermarking algorithm showed that the forgery detection accuracy of text document is improved from the average of 70% detection to 100% detection based on the content-based zero watermarking algorithm. The execution time of the proposed algorithm is calculated to be less than one second which can be concluded that the execution time performance is almost similar to other watermarking algorithm.