A Dual Authentication Watermarking Algorithm For Electronic Invoice Image With QR Code

Invoice is the original certificate of accounting and tax declaration. With the development of electronic commerce, paper-form invoice is not suitable for electronic commerce transaction which has the advantages of real-time and fast. The implementation of electronic invoice contributes to energy efficiency and environmental protection, regulating tax management, reducing operating costs, and improving operational efficiency. However, the phenomenon of fake invoices is becoming more and more serious. In order to realize the certification for electronic invoice image, an authentication watermarking algorithm is designed for electronic invoice image in this thesis.Firstly, the research background and significance are introduced. According to the characteristics of electronic invoice image, the authentication for it is divided into two parts: QR area and non QR code area. The main research content is the following two aspects:(1) The designed QR code image steganography algorithm is based on the following two characteristics:the ratio of black and white module ratio is close to 1:1; the decoding allows each module to have a certain range of deviations. Firstly, we select binary image as original image, which can make it easy to transmit and store via network. Secondly, adjust the boundary of the module within the permissible rang, which will ensure the carrier image be decoded correctly. Lastly, the embedding locations are selected randomly by a secret key, which make the algorithm safer. The experimental results show that the designed algorithm outperforms the existing algorithms in steganography capacity and in resisting scaling attack.(2) According to the characteristics of electronic invoice image, the format of GIF is chose. For the first level verification:important information in voice, such as the date and amount, is embedded into the QR part to achieve tamper detection. For the second level verification:since the color information of electronic invoice image is limited, we extend palette by rewriting, and then update index table, watermark is embedded by replacing the least-significant bit of the index value. This way makes the watermark totally invisible. To control increment of watermarked file, image is segmented and classified then into two categories, and just one category is considered to be recovered, besides, the recovery watermark for this category is generated by thresholding the image block. Then the recovery watermarks of one block are embedded into other two blocks to improve security. For situations where only need to certify the authenticity of invoice, we can just extract the watermarks in QR area to complete tamper detection, which can greatly improve the efficiency. If it needs to achieve tamper localization and recovery, we can use the second level verification.At the end, a simulation system is designed for all of the research achievements in this thesis.