Fingerprints have been used to authenticate people remotely and allow them access to a system. However, the fingerprint-capture sensor is cracked easily using false fingerprint features constructed from a glass surface. Fake fingerprints, which can be easily obtained by attackers, could cheat the system and this issue remains a challenge in fingerprint-based authentication systems. Thus, a mechanism that can validate the originality of fingerprint samples is desired. Watermarking techniques have been used to enhance the fingerprint-based authentication process, however, none of them have been found to satisfy genuine person verification requirements. This thesis focuses on improving the verification of the genuine fingerprint owner using watermarking techniques. Four research issues are being addressed to achieve the main aim of this thesis. The first research task was to embed watermark into fingerprint images collected from different angles. In verification systems, an acquired fingerprint image is compared with another image, which was stored in the database at the time of enrolment. The displacements and rotations of fingerprint images collected from different angles lead to different sets of minutiae. In this case, the fingerprint-based authentication system operates on the ‘close enough’ matching principle between samples and template. A rejection of genuine samples can occur erroneously in such cases. The process of embedding watermarks into fingerprint samples could make this worse by adding spurious minutiae or corrupting correct minutiae. Therefore, a watermarking method for fingerprint images collected from different angles is proposed. Second, embedding high payload of watermark into fingerprint image and preserving the features of the fingerprint from being affected by the embedded watermark is challenging. In this scenario, embedding multiple watermarks that can be used with fingerprint to authenticate the person is proposed. In the developed multi-watermarks schema, two watermark images of high payloads are embedded into fingerprints without significantly affecting minutiae. Third, the robustness of the watermarking approach against image processing operations is important. The implemented fingerprint watermarking algorithms have been proposed to verify the origin of the fingerprint image; however, they are vulnerable to several modes of image operations that can affect the security level of the authentication system. The embedded watermarks, and the fingerprint features that are used subsequently for authentication purposes, can be damaged. Therefore, the current study has evaluated in detail the robustness of the proposed watermarking methods to the most common image operations. Fourth, mobile biometrics are expected to link the genuine user to a claimed identity in ubiquitous applications, which is a great challenge. Touch-based sensors for capturing fingerprints have been incorporated into mobile phones for user identity authentication. However, an individual fake fingerprint cracking the sensor on the iPhone 5S is a warning that biometrics are only a representation of a person, and are not secure. To make thing worse, the ubiquity of mobile devices leaves much room for adversaries to clone, impersonate or fabricate fake biometric identities and/or mobile devices to defraud systems. Therefore, the integration of multiple identifiers for both the capturing device and its owner into one unique entity is proposed.
展开▼
