Article Preview
Top1. Introduction
Watermarking of images is a challenging research area which involves principles and techniques of communications, signal processing and encryption (Cox, Millar and Bloom, 2002). This research has many potential applications and is directed towards devising an efficient watermarking scheme which can be successfully implemented in spatial and / or transform domains. Robustness of embedding process is one of the important prerequisite of these schemes. Even if a scheme is proved to be robust enough to the selected image processing attacks, new attacks can be devised in future and thus, an apparently robust watermarking scheme can be proved fatal. The financial implications of copyright protection are enormous and there is a rigorous increase of research activity by famous research groups across the world at present. The research in the area of image watermarking in particular has taken two directions. The first one is oriented towards its SSP based industrial applications like the one discussed in this work. The second orientation is even more interesting and abstract in nature. It majorly involves the use of machine learning and other soft computing techniques to optimally embed the watermark into the optimally identified locations within the images or videos. A numerous research activity is carried out at present in this direction. Nearly all kinds of soft computing techniques- Artificial Neural Networks (ANNs) (Agarwal, 2013, 2015), Fuzzy Inference Systems (FIS) (Mishra, 2014, 2017), Support Vector Regression Models (SVMs) (Rajpal, 2016) as well as Meta heuristic techniques (GA and its variants) have been employed for this purpose. In addition to these techniques, hybrid intelligent networks are also profusely used to carry out watermark embedding and extraction. Therefore, it is sure that sooner or later the development of new and robust watermarking schemes will offer a competent and comprehensive solution to the problem of intellectual property rights (IPR) protection and content authentication. To achieve this objective, some steps should be taken. Such steps must consist of acceptance and implementation of a globally accepted watermarking protocol and devising watermark performance benchmarks. To this end, as a second line of research in signal processing and particularly the image processing in encrypted domain has attracted considerable attention of researchers in recent years (Erkin, Piva, Katzenbeisser, Lagendijk, Shokrollahi, Neven, and Barni, 2007). Although, cryptosystems play a significant task in secure signal processing, yet not all systems are suitable for this purpose. The main reason is that several cryptosystems like DES (Data Encryption Standard) and AES (Advanced Encryption Standard) do not preserve algebraic relations with the plaintexts after the encryption process (Elgamal, 1985). On the other hand, special cryptosystems such as those which offer homomorphism keep this relationship. Thus, they become suitable for signal processing applications.
The idea to carry out the proposed work emanates from the fact that in general, the encryption techniques are widely being used to enhance security in different signal processing applications. The problem of image watermarking is an important image processing application which actually deals with the issues of copyright violation and content authentication. As a result, encryption techniques are perceived as the most suitable tools to enhance security of this application. Paillier cryptosystem offers additive privacy homomorphism and is fast as required for a SSP application. It is therefore found to be suitable to carry out image watermarking. For this purpose, both the host image and the binary watermark are encrypted by using Paillier encryption technique. In the present work, prior to the encryption, the host image is compressed by using SPIHT compression technique.