Article Preview
TopIntroduction
With the enormous growth of the internet and the tremendous advancement in digital technology, the requirement of security measures is continuously increasing and has given rise to serious authentication and copyright issues (Erlich & Zviran, 2010). A wide variety of systems require reliable personal recognition which has led to the development of a number of biometric based personal authentication systems. Biometrics is the science of recognizing human beings uniquely based upon one or more intrinsic physical (e.g., fingerprint, iris, face, retina, etc.) or behavioral (e.g., gait, signature etc.) traits. Fingerprints are the most widely used parameter for personal identification amongst all biometrics. The reason behind this is their uniqueness and the fact that they do not change in the entire human life span (Maltoni, Maio, Jain, & Prabhakar, 2003). However, they are susceptible to accidental or intentional attacks when transmitted over a network. For this purpose, fingerprint images can be watermarked not only for preserving their fidelity but also to provide an authenticating mechanism to prove the validity of their owner. The combination of digital watermarking and biometrics is an emerging area for authenticating biometric data and securely transmitting it to various applications where it can be used for identification and classification purposes.
A number of watermarking techniques are available for embedding information securely in an image (Katzenbeisser & Petitcolas, 2002; Yang, Trifas, Francia, & Chen, 2009; Cheddad, Condell, Curran, & Mc Kevitt, 2010). In the past some data hiding techniques have been introduced particularly for protecting fingerprint data to increase the security level of fingerprint based systems, whereas there are others which hide data in any image for the purpose of secret communication or for copyright protection Data hiding techniques usually hide data in either of the two image domains (Gonzalez & Woods, 2005) i.e., the spatial domain and the transform domain. In the spatial domain watermarking, the simplest technique is to embed the data in the Least Significant Bits (LSB) of each pixel in the cover image. Transform domain techniques deal with embedding the message by modulating coefficients in the transform domain such as the Discrete Cosine Transform (DCT) or the Discrete Wavelet Transform (DWT) or the Discrete Fourier Transform (DFT).
Firstly, we discuss some existing watermarking techniques in the spatial domain. Uludag, Gunsel, and Ballan (2001, 2002) introduced a spatial domain technique for watermarking fingerprint images. The feature pixels or the feature regions of the host image were preserved in the watermarked image so as to facilitate its accurate fingerprint recognition at a later stage. They further extended their work to hide biometric data in fingerprint images (Jain, Uludag et al., 2002, 2003) where, a bit stream of biometric data (e.g., Eigen face coefficients, fingerprint minutiae data etc.) was embedded into selected fingerprint image pixels using a randomly generated secret key, so as to produce minimum distortion in the host image. Wu and Tsai (2003) presented a data hiding method in the spatial domain based on pixel value differencing. They used the differences of the gray values in the two pixel blocks of the cover images as features to cluster blocks into a number of categories of smoothness and contrast properties. Different amount of data could then be embedded in different categories according to the degree of smoothness and contrast. Chan and Heng (2004) further improved the quality of the stego image by proposing a method using an optimal pixel adjustment process. Particle Swarm Optimization (PSO) was used by Bajaj, Bedi, and Pal (2010) who proposed a high capacity data hiding scheme to embed a message in an image using LSB and by Bedi, Bansal, and Sehgal (2011) to hide an image within another image using LSB technique. These methods gave better results than the standard LSB technique and the techniques based on genetic algorithms and dynamic programming as the objective function used by PSO was based on the simple visual effect of the human visual perception capability.