지문 퍼지볼트 시스템의 VLSI 구현에 관한 연구

Abstract

Along with the development of internet, we could find and read the desired information anytime and anywhere we want. On the other hand, the outflow of personal information or crime using a blind spot of authentication system has been occurred frequently. Presently, the password method used in most of applications consists of the characters that can be easily memorized by user and the same password is used for several applications. Such password is possible for easy guess and outflow and has a risk of forgetting. Accordingly, the need for safer and more powerful authentication system is increasing. As the system that can make up the weakness of existing authentication system, the studies on the biometric system of using bio-information have been actively conducted. The fingerprint authentication of using fingerprint information is convenient and provides a powerful means of security at the same time. Since the fingerprint information change cannot be done freely due to the limited number of fingers, it may cause a serious problem if the fingerprint information saved for user authentication is exposed to other people. In order to protect the fingerprint information safely, the studied on the fuzzy fingerprint vault system have been conducted. The fuzzy fingerprint vault system generates a polynomial by using the secret keys of user and protects the fingerprint minutiae of user by constructing a fingerprint template together with the fingerprint minutiae after generating the chaff minutiae randomly. Although a few studies have been reported to protect the secret key of user as well as the fingerprint information of user by using the fuzzy fingerprint vault, they have a problem that cannot perform the alignment process due to the absence of fingerprint reference point. In order to solve this problem, the method to apply a geometric hashing technique to the fuzzy fingerprint vault has been proposed. The fingerprint alignment problem was resolved by performing a matching process upon comparing the enrollment hash table obtained through the geometric hashing technique with the verification hash table. However as a large number of chaff minutiae are added, the number of fingerprint templates is increased, thereby delaying the matching time from an increased number of hash tables generated. In order to reduce the execution time of fuzzy fingerprint vault system, this thesis proposes hardware architecture of fuzzy fingerprint vault system based on the geometric hashing. The proposed architecture is constructed by integrating the software and hardware modules. The hardware module consists of the matching module, polynomial restoration module, verification module, two memory for storing the enrollment hash table and verification hash table. The matching module compares the transformation minutiae of enrollment hash table with the transformation minutiae of verification hash table. The polynomial restoration module refers to the module that restores the polynomial coefficients by using the matched conversion minutiae. The verification module is the module that compares the coefficients of restored polynomial with the coefficients of the polynomial in the enrollment process. Also for the high-speed processing of fuzzy fingerprint vault system, this thesis proposes the hardware architecture that has applied a parallel processing technique. The hardware module consists of two memory for saving the enrollment hash table and the verification hash table, number of matching modules, and verification module. Based on the experimental results, the execution time in the matching process for 36 fingerprint minutiae, 200 chaff minutiae and 34 authentication fingerprint minutiae was 1.13 second when having implemented the system only with the software, 0.24 second with the proposed hardware architecture, and 0.01 second with the parallel processing hardware architecture of holding 64 matching modules.