Verification of iris image authenticity using fragile watermarking

Abstract

Abstract. This paper proposes and evaluates a watermarking-based approach to certify the authenticity of iris images when they are captured by a genuine equipment. In the proposed method, the iris images are secretly signed before being used in biometric processes, and the resulting signature is embedded into the JPEG carrier image in the DCT domain in a data-dependent way. Any alteration of the original (certified) image makes the signature no longer corresponding to this image and this change can be quickly identified at the receiver site. Hence, it is called fragile watermarking to differentiate this method from regular watermarking that should present some robustness against image alterations. There is no need to attach any auxiliary signature data, hence the existing, already standardized transmission channels and storage protocols may be used. The embedding procedure requires to remove some part of the original information. But, by using the BATH dataset comprising 32 000 iris images collected for 1 600 distinct eyes, we verify that the proposed alterations have no impact on iris recognition reliability, although statistically significant, small differences in genuine score distributions are observed when the watermark is embedded to both the enrollment and verification iris images. This is a unique evaluation of how the watermark embedding of digital signatures into the ISO CROPPED iris images (during the enrollment, verification or both) influences the reliability of a well-established, commercial iris recognition methodology. Without loss in generality, this approach is targeted to biometric-enabled ID documents that deploy iris data to authenticate the holder of the document.

Multi-Biometric Watermarking Scheme Based on Interactive Segmentation Process

“…The first subsection illustrates the experiments that have been conducted to evaluate the performance of the proposed segmentation algorithm (ISA) in comparison with the Hough transform based segmentation [21][22][23][24]. The second subsection presents a general comparison between the proposed MBW scheme and the schemes in [3,9,[11][12][13][14][15][16][17].…”

“…Table 5 shows a general comparison between the proposed MBW scheme and the schemes in [3,9,[11][12][13][14][15][16][17]; where the schemes [11][12][13][14][15][16][17] have been applied to iris images in the study that have been presented in [8], therefore, the comparison here depends on the performance of these schemes in [8]. As shown in Table 5, the proposed scheme performs better than all the compared schemes in [3,9,[11][12][13][14][15][16][17] in terms of intactness of the iris region which is an essential property to avoid affecting the features of the iris region and consequently there will be no effect on the recognition performance. In terms of robustness, the proposed scheme performs better than the schemes in [9,[11][12][13][14][15][16][17].…”

“…Iris-based biometric systems are highly recommended because of their security and ease of use [4], therefore, a number of watermarking schemes have been presented to protect iris images either by extracting the template of the iris image and embedding it in a cover image [5][6][7] or by embedding a watermark and secret data in the iris image [3,8,9]. In [10], a study has been conducted to investigate the effect of the two scenarios that are mentioned above and it proved that, the second scenario in which the watermark is embedded in the iris image obtained better results in recognition performance, therefore, the research in this paper will focus on the second scenario.…”

“…The study proved that the watermarking process reduces the images' recognition performance significantly when the payload is high. In [9], another approach has been presented which depends on applying a fragile watermarking to embed a secret signature in the DCT coefficients of the iris image. Any change in the iris image will destroy the signature and thus the alteration can be easily detected.…”

“…The robust watermarking technique in [3] suffers from the limited capacity as well as the distortions that are presented in the iris region because of the watermarking process. In the fragile watermarking techniques [9,[11][12][13][14][15][16][17], the watermark does not have robustness against attacks and thus only standardized transmission channels and storage protocols can be used. In addition, the process of embedding the watermarks in an iris image will cause distortions in the iris region.…”