Research On Template Protection Method Applied To Iris Recognition

With the rapid development of information technology,more and more fields require reliable and secure identity authentication technology.Biometric recognition has developed rapidly driven by various market demands,and iris recognition technology has become one of the popular researches due to its many advantages.However,with the application and development of iris recognition technology,some security risks has been gradually exposed.Once the iris feature database is stolen,it will lead to the leakage of user information and affect the safety of the user's personal property.Therefore,the original data protection for iris features has become one of the main research hotspots in this field.In this thesis,the main research content is the method of the iris feature template protection in the iris recognition system.The preprocessing,feature extraction,template protection and matching methods of the iris image is studied,and a dual iris template protection method is proposed.In the iris image pre-processing stage,a classic algorithm is used for iris localization and normalization.Then the iris texture features are extracted with Log-Gabor transform.In order to protect the original iris feature data obtained,a dual iris template protection method based on local scrambling and double random phase encoding?DRPE,Double Random Phase Encoding?is proposed in this thesis.Both irises are needed in this method.The left one is used as the verification iris,and the right iris's feature vector is used as the key.First,the feature matrix of the user's left iris image is scrambled using Arnold blocks to obtain the first protection of the iris feature code?IrisCode?.Then,the IrisCode of the other iris of the same user is extracted,and it is used as the second phase mask of double random phase encoding.With this key,a new template is generated in the fractional Fourier domain by double random phase encoding and stored in the database to achieve the purpose of protecting the iris template.In the matching stage,this thesis uses Hamming distance to classify the matching results.In order to illustrate the applicability and universality of the method in this thesis,the three sub-libraries of the CASIA-IrisV3 dataset,?namely CASIA-IrisV3-Interval,CASIA-IrisV3-Lamp,CASIA-IrisV3-Twins?and MMU-V1 dataset,were used for simulation experiments and performance evaluation.50%samples of each person in each gallery are randomly selected as the training set,and the rest as the test set.The final correct recognition rates of the algorithm in this thesis in the four datasets respectively are 98.73%,83.35%,78.99%and 73.65%.These recognition accuracy are slightly reduced compared with the unencrypted method,but the algorithm trades for a better template protection effect at the cost of a slightly lower recognition accuracy.The decidability of the algorithm proposed is high,and they are 5.64,5.02,5.97 and4.87 respectively of each dataset.It indicates that The distance between the real distribution the impostor is large,and proposed method improves the performance of template protection.In addition,²₅₂ kind of templates can be generated by the proposed algorithm,and the attacker needs to try₃₁₄^52?624?_{4 2}times to restore the original data.It is shown that the proposed algorithm can satisfy the three criteria of template protection,namely irreversibility,revocability and unlinkability.This method can also avoid mountain climbing attacks and related attacks.In summary,the proposed method achieves the protection of the original feature data of the iris while satisfying a higher recognition rate.