A Fingerprint Templates Protection Scheme Based On Double Rotation Feature Descriptor

With the rapid development of computer technology,information security technology and machine learning technology,biometric-based identification technology has been rapidly developed.Currently,identity authentication using fingerprints is widely applied in access control systems,electronic product unlocking and the personal financial payment field.Thus,user fingerprints and other biometric information are increasingly associated with personal property.The fingerprint is an inherent biological characteristics of every individual.Once lost,it will be lost forever,and it will cause enormous harm to person.Therefore,the identity authentication system based on fingerprint identification should have high accuracy of identification while ensuring the safety of users' original fingerprint.How to ensure user fingerprint information security is one of the key issues for the breakthrough of fingerprint recognition to the next large-scale application.Therefore,it has attracted a lot of scholars' attention and great efforts have been made to the research of fingerprint template protection.In the process of fingerprint template protection,due to the influence of systematic deformation and elastic deformation,the acquisition of fingerprint images inevitably introduces certain errors.Therefore,irreversible encryption of the user's original biologic features always has a certain degree of inhibition modification,which makes it unable to tolerate the existence of error and leads to the lack of accuracy in matching the encrypted domain of the fingerprint template.In view of the above problems,this thesis proposes a fingerprint template protection algorithm based on double rotation feature descriptors,and also carries out some innovative work in the field of fingerprint template protection.The main contents include the following aspects:(1)In this thesis,a composite field minutiae descriptor based on the directional field and frequency field is used,which extends data from traditional descriptors.This data augmentation enriches the fingerprint feature information and improves the accuracy of fingerprint matching.At the same time,it can combine a longer key to improve the security strength of fingerprint encryption.(2)In this thesis,a new minutiae triangles operator based on alignment-free is proposed.This operator has rotational invariance compared with the traditional minutiae triangles operator,which avoids the adverse effect on the final matching result due to the imprecision of the fingerprint registration.(3)This thesis proposes a ridge counting algorithm based on soft interval,which takes full account of the problem of elastic deformation of the fingerprint.The counting method of the ridge line is optimized by adding elastic variables to improve the counting accuracy of the ridge line.(4)This thesis uses an elastic coding and quantification method that allows certain error for the pre-encoded biometrics.Combined with the local DFT irreversible encryption function,a more secure and reliable fingerprint template protection method is realized while ensuring the recognition accuracy.(5)This thesis implements user account management system based on fingerprint encryption,to solve the problem of inconvenience of user account management.The user can realize automatic login account to use their fingerprints.The fingerprint template protection algorithm based on double rotation feature descriptors is verified on the public fingerprint database FVC2002 DB1 and DB2,and the EER(equivalence error rate)is 3.16% and 2.31% respectively.Experiments show that the recognition performance of the fingerprint template protection algorithm is superior to the traditional classical algorithm,and the detailed security performance analysis is given in this paper.