Digital images authentication scheme based on bimodal biometric watermarking in an independent domain

https://doi.org/10.1016/j.jvcir.2016.02.006Get rights and content

Highlights

  • We propose a new biometric-based bimodal watermarking scheme.

  • Two independent watermarks based on fingerprint and iris biometrics are used.

  • Our algorithm explores an independent domain of the cover image.

  • Authentication procedure is based on fusion of two extracted biometric watermarks.

Abstract

With the growing accessibility and usability of internet there is a growing concern over content protection of digital images. Recently, to eliminate the traditional use of passwords and to ensure that the access to the image is restricted only to legitimate users, security solutions are increasingly combined with biometrics. Consequently, biometric-based watermarking algorithms, that involve embedding the identity of the owner, are proposed to solve ownership disputes. This paper presents a new scheme for protecting and authenticating invisibly watermarked digital images. It applies Independent Component Analysis to the cover image and enables the insertion of two independent watermarks based on fingerprint and iris biometrics. In this approach biometric techniques are used for watermarks generation and for owners authentication. The main advantage of proposed algorithm is construction of ICA based watermarking domain to enable insertion of two independent watermarks, that improve authentication accuracy and makes scheme more robust.

Introduction

Nowadays unrestricted access to computers and computer networks makes it possible to share multimedia data such as digital images among many people, not only authorized users. To ensure that images are transmitted from the correct sources to the intended recipients (authentication) many systems, including cryptography, encryption, steganography and watermarking techniques, have been elaborated, of which some examples can be found in [1], [2], [3], [4]. Recently, these security technologies are increasingly combined with biometric techniques mostly due to its ability to differentiate between authorized and unauthorized users (e.g. [5], [6]). As a result, biometric-based security schemes, using unique, measurable physiological and behavioral characteristics of a users, represent a robust alternative to traditional copyright protection and authentication schemes. Many examples of such systems combining biometric features with encryption [7], cryptography [8], steganography and watermarking [9], [10] have been provided so far. As a continuation of these studies, this paper is concerned with overcoming the major limitation of traditional watermarking schemes, that is lack of watermark uniqueness, introducing biometric-based watermarks.

In general, watermarking technology, that is subject to the same security requirements as cryptography ([2]), solves important security problems including copyright protection and authentication (e.g. [1], [11]). In such schemes watermark data are hidden in the image, so that they could identify the owner during potential ownership controversies. However, as traditional watermarks such as logo, text or images, do not have a user based uniqueness, they may lead to some imitation, tamper and repudiation. Such watermarks could be hacked by the hackers (e.g. at brute-force attack) and use by them to claim ownership or the host image could be fraudulently watermarked by impersonators. Therefore, biometric-based watermarks, that are generated using biometric traits of the image owners, become solution to these issues, as they provide the requisite level of authentication. Employment of user-specific watermarks is critical for determining if certain individual is authorized to access the content or whether current user is a legitimate owner of the work. In case of any ownership dispute on the image, the watermark is extracted from the image and compared against different samples of the person claiming ownership. If they match, the claiming person is the actual owner of the image.

Several methods of using watermarking techniques in conjunction with biometrics have been proposed in the literature. Most of them adopt the idea of hiding one biometrics in another biometric image to protect the originality of biometric templates stored in databases, e.g. [12], [13]. Although there are also methods, that this paper is concerned with, where biometric data are inserted into neutral images, sometimes even unrelated to this data, so that after extraction could be used for copyright protection purposes. For example [14] presents method of embedding fingerprint data in a watermarking domain, that is combination of DWT and SVD transformations. Authors proposed to use minutiae coordinates matrix after SVD transformation as a watermark and after extraction compare it with data provided by user claimed ownership. Another algorithm in which binary iris code is inserted in the neutral image is presented in [15]. In this scheme decision about user acceptance depends on similarity measure between extracted and provided by users biometric samples, that is computed as ratio of correctly detected watermark bits. In this field there are also methods, that use more than one biometric watermark, mostly to take advantages of multimodality, e.g. [16], [17], [18]. In [16] authors present method of embedding face, speech and signature data at different stages of watermarking procedure, that combines different transformations, e.g. DWT, DCT and Linear Predictive Coding (LPC). After extraction biometric traits are separately compared with their original versions using correlation and similarity factors, but there is no authentication procedure that fuses these biometrics. Next two methods presented in [17], [18] deal with embedding fingerprint and iris features as the binary watermarks in DCT based watermarking domain. In [17] features of both traits are used to construct one watermark (fusion at feature level), that after extraction is again divided into two biometrics. Then both traits are compared with their original versions separately, using normalized correlation for iris and minutiae matching procedure for fingerprints. Finally, in [18], where two binary watermarks are inserted modifying the amplitude of some selected coefficients of the cover image in DCT domain, user authentication procedure is performed after fusion of both biometric traits at the decision level. Proposed algorithm is tested for real biometric databases and considering attacks influence to demonstrate advantages of multimodal authentication procedure.

This paper proposes a new bimodal biometric watermarking algorithm based on Independent Component Analysis (ICA) of the cover image, that enables the authentication of digital images owners. In our algorithm two biometric watermarks based on fingerprint image and binary iris template are embedded invisibly in a neutral cover image. After watermarks extraction, instead original watermarks, just another samples of biometric traits are required to perform owners authentication. Our scheme involves ICA decomposition of the cover image into four basic components and choosing two of them, with the highest energy ones, for embedding. The first component (approximation of the cover image) is mixed with scrambled fingerprint image, whereas in the second component (details of the cover image) binary iris template and its mask are embedded by pixels quantization. The main reason for applying ICA in this algorithm is its ability to decompose any image into statistically independent source images. In that the cover data is partitioned into disjoint components and each watermark is embedded into them independently. Incorporating idea of using two independent biometrics make system robust to variations of biometric data, processing connected with watermarking and influence of possible attacks.

The novelty of the proposed method is introducing Independent Component Analysis (ICA) to biometric watermarking field, as to the authors knowledge, it is the first biometric watermarking algorithm, that is based on ICA transform. In our method standard ICA approach, that involves embedding only one image watermark, is employed and extended, so that insertion of two biometric watermarks would be possible. Moreover, embedding procedure is adjusted to the watermarks as selection of ICA components for embedding depends on representation of biometric watermarks. Finally, we propose owners authentication procedure based on biometric data fusion at matching score level, that improves user authentication accuracy even when watermarked images undergo some attacks. The main goal of our simulation study is to elaborate how user verification accuracy changes after watermarking procedure and how significant improvement of these results could be achieved when information provided by both biometrics is fused.

The paper is structured as follows. Section 2 introduces watermarking in ICA domain bases and some issues connected with biometric verification methods that will be elaborated in our scheme. Proposed methodology is described in Section 3. The experimental study, algorithms settings and results discussion are presented in Section 4. Section 5 concludes on the study carried out.

Section snippets

Preliminary

In this section, we give an introduction to the properties of standard watermarking algorithm in ICA domain, i.e. MRICA method, that will be extended to construct watermarking domain in our scheme. Then some information about biometric verification methods, especially based on fingerprint and iris images, that will be used for watermarks construction, are provided. Finally, some issues connected with biometric verification errors and multimodal biometric systems are discussed.

The proposed bimodal biometric watermarking scheme

In this section the embedding and extraction stages of the proposed watermarking scheme are presented, emphasizing the novelty of our algorithm compared to standard MRICA method, that was described in Section 2.1. Then the authentication procedure, based on two extracted watermarks and incorporating the idea of biometrics fusion, is provided.

Experimental results and discussion

In this section a multimodal biometric database, including fingerprint and iris images, that is used in the simulation study, is first described. Then experiment settings are provided in reference to particular stages of the proposed watermarking scheme and authentication procedure. Finally, results of user verification based on the extracted watermarks for single as well as fused biometrics are presented and discussed.

All computations were executed on a personal computer (Intel Core i7 2.20 

Conclusions

In this paper we proposed authentication scheme to protect the ownership of digital images using a bimodal biometric watermarking approach. Presented methodology is based on incorporating fingerprint and iris biometric verification methods for construction of biometric watermarks and embedding them in an independent domain of the cover image. The performance of proposed algorithm, also including some processing on watermarked images, was evaluated in terms of user authentication errors. The

Acknowledgements

Research of Wioletta Wójtowicz was supported by the program “Doctus – Małopolski fundusz stypendialny dla doktorantów”.

We would like to acknowledge the Group of Machine Learning and Applications from Shandong University for providing the SDUMLA-HMT database.

References (36)

  • A. Jain et al.

    Hiding biometric data

    Trans. Pattern Anal. Machine Intell.

    (2003)
  • W. Wójtowicz et al.

    Biometric watermarks based on face recognition methods for authentication of digital images

    Security Commun. Netw.

    (2015)
  • N. Komninos, T. Dimitriou, Protecting biometric templates with image watermarking techniques, in: Proceedings of...
  • M. Vasta et al.

    Feature based rdwt watermarking for multimodal biometric system

    Image Vision Comput.

    (2009)
  • N. Rao et al.

    A novel scheme for digital rights management of images using biometrics

    Int. J. Comput. Sci. Netw. Security

    (2009)
  • A. Hassanien

    Hiding iris data for authentication of digital images using wavelet theory

    Patten Recogn. Image Anal.

    (2007)
  • V. Inamdar et al.

    Dual watermarking technique with multiple biometric watermarks

    Sadhana

    (2014)
  • M. Dutta, Generation of digital signature from multi-feature biometric traits for digital right management control, in:...
  • Cited by (34)

    • Securing visual search queries in ubiquitous scenarios empowered by smart personal devices

      2020, Information Sciences
      Citation Excerpt :

      Again, we remark that the blocks represented in Fig. 2 are the minimum building elements necessary to maintain the interoperability. Digital watermarking techniques are widely used in several applications [1,2,6,14,17,21], including copyright protection, integrity and many other security aspects. As addressed in [15,16], most of the watermarking schemes operate essentially in two domains:

    • A novel Frei-Chen based fragile watermarking method for authentication of an image

      2022, Concurrency and Computation: Practice and Experience
    • Human-AI Protocols for Cloud Data Management

      2022, Lecture Notes in Networks and Systems
    View all citing articles on Scopus

    This paper has been recommended for acceptance by M.T. Sun.

    View full text