Compressed encrypted domain watermarking and forensics for media security

• Main Author: Venkata, Subramanyam
• Other Authors: Sabu Emmanuel
• Format: Theses and Dissertations
• Language: English
• Published: 2013
• Subjects: DRNTU::Engineering::Computer science and engineering
• Online Access: http://hdl.handle.net/10356/52175
• Institution: Nanyang Technological University

Digital media content creation/capturing, processing and distribution have witnessed a phenomenal growth over the past decade. The ease with which this digital media can be edited has also led to the tremendous growth in illegitimate usage and distribution. Digital Watermarking and Forensic techniques emerged as possible solutions to track and minimize such illegitimate activities. The focus of this research study is to design the compressed-encrypted domain robust image and audio watermarking techniques for purposes such as proof of ownership, copyright violation detection and fingerprinting. In addition, we focus on forensic algorithms for establishing video authenticity. The digital media is often managed in compressed-encrypted format, for example in Digital Rights Management (DRM) systems, and sometimes it is necessary to watermark in this compressed-encrypted domain itself. The compressed-encrypted domain media refers to the content which is compressed first and then encrypted using a secure cryptosystem. Different compression techniques like JPEG2000, JPEG-LS or JPEG for image and MP3 for audio, are generally used. It is a challenge to watermark the compressed stream as a small modification in the compressed data may lead to a considerable deterioration in the quality of decoded image. Further in an encrypted piece of content, changing even a single bit may lead to random decrypted values. It should also be possible to detect the watermark correctly even after the content is decrypted. And the compression gain should not be significantly compromised as encryption may lead to cipher text expansion. The first part of our research is in compressed-encrypted domain robust image and audio watermarking. Towards this, we proposed a watermarking algorithm for JPEG2000 compressed images based on Spread Spectrum (SS) watermarking technique and homomorphic RC4 stream cipher for the purpose. In this proposed scheme, embedding is done in compressed-encrypted byte stream while detection can be performed in compressedencrypted, decrypted or decompressed domains. The benefit of using the RC4 cryptosystem is that it is secure, does not lead to cipher text expansion and allows watermarking in a predictable manner. Further, to enhance the embedding capacity, we investigate different watermarking techniques such as Scalar Costa Scheme Quantization Index Modulation (SCS-QIM) and Rational Dither Modulation (RDM). We perform elaborate theoretical and experimental evaluation to demonstrate the effectiveness of the proposed algorithm. Different compression schemes differ in their nature of pipelines. Therefore, a unique algorithm is not applicable for all the compression schemes. Towards this, we proposed a watermarking scheme for compressed-encrypted JPEG-LS images. In this scheme the encoded bit stream is encrypted using a secure stream cipher while context is used for embedding watermark. Detection is performed in decrypted domain. Watermarking using context does not degrade the image quality which is often desired in biomedical images. We study in detail the embedding capacity, robustness, perceptual quality and security of the proposed scheme. In the next part of our research, we proposed a robust JPEG image watermarking scheme in partially compressed-encrypted domain based on SS watermarking technique. While the proposed technique embeds watermark in the partially compressed-encrypted domain, the extraction of watermark can be done in the encrypted or decrypted domains. The advantage of the proposed scheme is that, it not only facilitates partial compressedencrypted domain watermarking but also minimizes the reduction in compression efficiency. In addition, we propose a technique to improve the embedding capacity by using coefficient-pair difference. The embedding capacity increases due to the fact that the proposed technique removes the host signal interference. Further, we present an application of the proposed watermarking technique to watermark partially compressed-encrypted MP3 audio. We discuss the effectiveness of the proposed algorithm using analytical and experimental results, and compare against the existing encrypted domain algorithms. Media authenticity is another crucial aspect of media security. As a second part of our research work, we designed a forensic technique for verification of the integrity of the video for copy-paste forgery. However, the nature of encoding - intra and inter modes as well as the complexity of tampering, makes it very challenging to identify a video forgery. The proposed scheme is based on Histogram of Oriented Gradients (HOG) and video compression properties. The proposed algorithm can detect the forged regions under wide variety of signal processing manipulations such as compression, filtering, or scaling. Further, to detect general forms of tampering such as frame - editing, addition, deletion, reordering, duplication, we proposed a holistic video tamper detection technique based on estimation principles and double compression. The advantage of this algorithm is that it can detect tampering of I, P or B frames in a GOP with high accuracy. Experimental results establish the effectiveness of the proposed algorithm.