2004 IEEE lntemational Conference on Systems, Man and Cybernetics Digital Watermarking based Secure Multimodal Biometric System* Mayank Vatsa, Richa Singh, P. Mitra Department of Computer Science & Engineering Indian Institute of Technology Kanpur, INDIA mayanl-richa@yahoo.com, pmitra@iitk. ac.in Abstract - This paper presents o multimodal biometrics gwtem using watermarking algorithms with hvo levels of security for simultaneously verifying on individual and protecting the biometric template. Iris templatc is Watermarked in face, such that the foce is visible for verification ond the wotermorked iris is used to cross authenficate the individual and secure the biometrics data os well. The accurocy of the multimodol biometrics system is around 96.8%. This system is olso resistant to common attacks on biometric templates. Keywords: Watermarking, Multimodal Biometrics, Face, Iris, Radial Basis Function. 1 Introduction In recent years there has been an explosive growth of business-to-customer activities over the Intemet. The total value of these web-based transactions is over several billion dollars. At present, buyers are authenticated by service providers using a combination of user ID and password. The critical information about the transaction, such as the credit card number and amount, are sent over the web using secure encryption methods. However, current systems are not capable of assuring that the transaction was initiated. by the rightful owner of the credit card. As Internet revenues grow, credit card owners and credit card issuers are likely to he increasingly concerned with the reliability :md security of transactions. To solve this problem, bionietric techniques offer reliable methods for personal verification; but the problem of security and integrity of the biometrics data poses new problems. If a person’s biometric data is stolen, it is not possible to replace it as in the case of a stolen credit card, ID or password. In order to promote the wide sprcad utilization of biometric techniques, an increase in the security level of biometric data is necessary [l]. Encryption and watermarking are among the possible techniques to achieve this, but encryption does not provide security once the data is decrypted. Watermarking involves embedding information into the host data to provide geater security. Since embedding the watermark may change the inherent characteristics of the host image, verification performance Afzel Noore Lane Department of Computer Science & Electrical Engineering West Virginia University, Morgantown, USA afzel.noore@mail.vwu.edu based on (decoded) watermarked images should not be inferior to the original non-watermarked images. Recent work has been undertaken on watermarking of fingerprint images [3, 41. Fingerprint images or minutiae (distinguishing characteristics of fingerprint images) are hidden in images. Jain et al. [6] have proposed an effective way of hiding Eigen faces in fingerprint images. These are the only works on biometric data hiding and these involve the use face and fmgerpiint images. In this paper, two levels of security are proposed for simultaneously verifylng any individual and protecting the biometric template. One biometrics template i.e. iris is watermarked in face, such that the face is visible for verification and the watermarked iris is used to cross authenticate the individual and secure the biometrics data (face) as well. For generating the watermark data i.e. iris template, an algorithm based ou 1D log Gabor is used. 1D log Gabor is convolved with the transformed texture image of iris and thus the template is generated. This template called as iriscode is in the binary form and is unique for every individual. It is now embedded into the face image of the same individual to protect the face template as well as perform the multimodal operation. For watermarking, two algorithms namely Modified Correlation based algorithm and Modified 2D Discrete Cosine Transform based algorithm are presented. 2 Choice of Watermark Object The first question we need to ask with any watermarking based multimodal biometric system, is what form will the embedded watermark take and which biometric template should be used as the watermark? Previous work on hiding face in fingerprint images [6] and hiding fingerprint minutiae in an image [3, 41 are designed to hide the extracted features of the watermark image. These algorithms allow an image to cany information in the form of features. The drawback however with these approaches is that by compressing the watermark-object before insertion, robustness suffers. So rather than using feature characters as watermark, we propose using the binary image as watermark. * 0-7803-8566-7/04/$20.00 0 2004 IEEE. 2983