Research On The Fingerprint Identification And The Fuzzy Vault Algorithm Based On The Least Square Approximation

The fingerprint identification technique, as a kind of identification techniques, is widelyapplied in the information security field. However, the low-quality fingerprints are not good for thefingerprint identification technique. Hence, it is meaningful to deal with the low-qualityfingerprints using the fingerprint preprocessing technique. The fingerprint preprocessing techniquesinclude the fingerprint orientation calculation. Therefore, improving the fingerprint orientationalgorithm is of much value.The fuzzy vault algorithms based on the fingerprint features are a kind of biometric encryptiontechniques and they can protect the fingerprint feature template and the key or secret information.They combine the fingerprint features with the key, which make that they are less vulnerable tosome attacks. On the basis of the fuzzy vault algorithm proposed by Juels and Sudan, we proposetwo kinds of improved fuzzy vault algorithms based on fingerprint features, which are moreeffective in resisting the based attack, the brute attack and the correlation attack.The major innovative points in this paper can be described in the following three parts:1. We implement the fingerprint identification based on the least square approximation.Especially, we propose an adaptively modeling fingerprint orientation algorithm based on theLegendre polynomial and the least square approximation. We stimulate the proposed algorithmusing the FVC2002-DB1, and the experimental results show that our algorithms better represent thefingerprint structure than the algorithm proposed by Ram etc al. Meanwhile, by further stimulation,our algorithm shows more effective than the algorithm proposed by Ram et al. in both thefingerprint enhancement based on Gabor filter and the fingerprint minutiae-based matching .2. We proposed a new fuzzy vault algorithm based on the password and Chinese remaindertheorem. Based on the fuzzy vault algorithm given by Juels and Sudan, we use the password toencrypt the random key, and the encrypted key is used as the coefficients of a constructedpolynomial. We project all minutiae onto the polynomial and use the chaff point placementtechnique proposed by Orencik et al. to prevent the existed rules between the minutiae and the chaffpoints. In the vault decoding, we use the Lagrange interpolation method and Chinese remaindertheory to recover the key. Compared with the fuzzy vault scheme proposed by Nandakumar et al.,our algorithm is more effective in resisting the based attack, the brute attack and the correlationattack to some extent. Our simulation in Matlab shows that the GAR (the genuine acceptance rate)is not decreased and the FAR (the false acceptance rate) is not increased in our scheme.3. We proposed a novel fuzzy vault algorithm based on fingerprint features and rectangular grids. In our algorithm, all minutiae are supposed to be within a circle with its center being aminutia, and they are projected onto some rectangular grids. If a grid includes some minutiae, it isfilled with 1, otherwise, it is filled with 0. Afterwards, a bit-string with a certain length is generatedby sequentially visiting the grids in the rectangular grids and then stored as the abscissa of thevault’s point. It makes that an attacker is difficult to get the fingerprint features information, even ifthe attacker can obtain some genuine points. Compared with the previous fuzzy vault algorithmproposed by us, our algorithm is much more effective in resisting the based attack, the brute attackand the correlation attack to some extent. Our stimulation in Matlab shows that the GAR isincreased and the FAR is decreased to some extent in the novel algorithm.