论文部分内容阅读
With the large-scale proliferation of biometrics-based personal authenticationsystems, its security and privacy concerns have emerged as a hot research issue. A lotof researches have been conducted on the pattern recognition area of biometrics regime,unfortunately; the critical security and privacy issues to attain the reliable andtrustworthy identity authentication are being ignored. A biometrics system, even withsignificant recognition performance, is only universally acceptable if it is secure,reliable, robust against attacks and threats, and protects a users privacy in efficientmanner.
To overcome the biometrics security challenges and to alleviate the biometricsprivacy issues, in this thesis, we investigate the methods and techniques aiming atthwarting the aforementioned problems inherent in biometrics systems.
For the secure transmission of biometrics data over insecure network channel, twodifferent modes of data hiding schemes are proposed. The first method, which isimage-based hidden transmission of biometrics template, hides the biometric data intoa group of different images. Secret keys are generated by the biometric image andencryption is applied on the biometric templates before hiding into the host images.Decoding performance of hidden templates is encouraging and outperforms in the caseof transmission attacks on the cover image. The second proposed method is an audiosignal-based covert-transmission scheme of biometric templates, in which we used anovel chaos and NDFT-based template hiding scheme, and proved that the combinationof encryption, encoding, modulation, and data hiding of biometric template in audiosignal could not affect the identification performance of the biometric recognitionsystem. The biometric data extraction performance of the system is efficiently capableof identifying a person without stringent compromise of matching accuracy.
To obstruct the non-liveness and retransmission issues of biometrics images, achallenge/response-based biometrics image scrambling system and method is proposed.In contrast to general challenge/response systems, which issue challenges to the user,our scheme is based on a novel architecture of intelligent biometrics sensor that hascomputational power to receive challenges from the authentication server and generateresponse to the challenge with the encrypted biometric image. The encryption anddecryption performance of the proposed system is encouraging to use in the real-timesystems of biometrics image transmission.
In order to strengthen the privacy of biometrics data, an efficient and tokenlesstechnique is presented. Firstly, this scheme extracts features from the capturedbiometrics face images to generate the face feature vector and then, pseudorandomnumbers without using the tokens e.g. USB or smart card are generated, which can becombined with the face feature set to attain a binary FaceHash code. This schemeeliminates the fear of using biometrics for the cross matching applications, andprovides a remedy for non-revocable biometrics data.
Finally, an efficient and practical chaotic hash-based fingerprint biometric remoteuser authentication scheme on mobile devices, e.g. cell phone and PDA, is investigated.This scheme is completely based on one-way collision free chaotic hash functions, anddoes not maintain password tables and biometrics database on the remote server.Furthermore, users can choose their passwords freely and change or update themsecurely whenever they want. This scheme is encouraging and better than using thelarge exponential functions or discrete logarithm problems on mobile devices, becauseof their low computational power and speed.
Hence, the contribution of this thesis is envisaged in enhancing the usersconfidence and trust to use biometrics systems without having a fear of security breachand privacy compromise for the depen.dable and trustworthy personal authentication.