Development of Copyright Protection and Authentication Schemes with Extended Visual Cryptography

Abstract

The existing image watermarking schemes protect the digital image by embedding watermark that degrades the visual quality of the image. This embedded watermark is extracted to prove the copyright or authentication. Visual Cryptography based watermarking schemes have gained a lot of attention in recent years since they aid in protecting the images without modifying them. The focus of this thesis is on two key uses of watermarking, namely copyright protection and authentication. Most of the existing Visual Cryptography based image authentication schemes have high computational cost as they embed the constructed shares and authentication data into the cover images. Most of these schemes create expanded and noisy shares, leading to pixel expansion and security threat of some sensitive information being shared, respectively. To address these problems, an imperceptible authentication scheme is introduced for grey images where meaningful and un-expanded authentication shares are generated from watermark and cover images. Cellular Automata is used to construct the master share that provides self-construction capacity to the share, thereby saving the storage cost and enhancing stability. The watermark is recovered just by superimposing these authentication shares, thereby helping to reduce the computational cost at receiver’s side. The scheme possesses the ability of tamper detection, localization and lossless recovery as of the tampered data as well. The copyright protection schemes are proposed to protect the ownership of the digital images. Most of the existing copyright protection schemes based on visual cryptography produce noisy shares that is a security threat. For this purpose, a robust and secure copyright protection scheme based on curvelet transform, K-means clustering and extended visual cryptography for color images is proposed that creates meaningful noiseless shares. The scheme is robust to withstand a range of image manipulation attacks and provide greater imperceptibility. The effectiveness of the scheme is shown by comparing it with existing copyright protection schemes. However this scheme was unable to handle false positive cases, where the false owners can also prove their copyright on the image. As a result, this scheme has been enhanced to cope successfully with false positive scenarios by using two watermarks. One of them is provided by the user and the other is created from the host image. The latter is used to cope with false positives as it contains the host image information. This scheme works only for single images with single owners. As a result, this scheme was further extended to multiple images and multiple owners. To the best of our knowledge, two copyright protection schemes (Amiri and Moghaddam(2016), Liu and Wu (2011)) exist in the literature, for multiple images. These schemes have some limitations like meaningless shares, false positives and side information transmission. In addition, these schemes involve the presence of every ownership share to prove the copyright. This limits the schemes, as if any of these ownership shares are not accessible, the other owners can not prove their copyright. To fix these concerns, a robust copyright protection scheme for multiple color images with multiple owners is proposed that creates meaningful shares where a qualified set of owners can verify the ownership. Three types of shares are included in this scheme: i.e. master share, ownership share and key share. The ownership share is generated using a master share and a watermark that is then used to construct a key share that is stored with the Trusting Authority. To prove the copyright of multiple images, the ownership shares and the key share are superimposed to retrieve the watermark. The visual appearance of the watermark proves the copyright of rightful owners. Experimental results demonstrate the efficiency of the proposed schemes. Comparisons with existing Visual Cryptography based watermarking schemes show improved performance of the proposed schemes.