| Summary: | 碩士 === 國立中山大學 === 通訊工程研究所 === 107 === This study proposes a biometric system for personal identification based on the biometric characteristics from the iris image. A protection scheme with a high security level is applied to the biometric template data to guarantee its revocability, security and diversity. Error-correcting codes (ECC), cryptographic hash functions (CHF), and the proposed Template Reliability Mapping and Dominating Feature Points technologies are the core of the devised template protection scheme. The topics addressed in this thesis include: (i) Extraction of Soft Reliability Value, which is the key that leads to the success of the next two technologies. (ii) Template Reliability Mapping, which allows us to easily manipulate the error correcting capability for the identification system by using the same error correcting code. Different from the cases in conventional communication systems, the error correcting code with a higher level of error correcting capability may not be the best choice for the considered identification system. Therefore, it is very important to be able to freely adapt the error correcting capability. In contrast, in the conventional schemes, The error correcting capability can only enhanced or reduced by replacing different codes. This technology also allows us to add Hash function to the iris recognition system to improve system security while retaining the system performance. (iii) Dominating Feature Points, which allows us to overcome the limitations of the iris identification data, where the parity bits used in the error correcting code is always greater than the message bits. Consequently, we are able to significantly enhance the security level and speed up the identification process. Moreover, the original iris data is able to provide the property of substitutability to further enhance the system security level. Lastly, it has been verified that the proposed system achieves promising recognition and security levels, when considering publicly available iris image databases.
|
|---|
