Research And Implement Of The Privacy-preserving Iris Recognition

Recent advances in biometric recognition make the use of biometric data more prevalent for authentication and other purposes.Due to its advantages of high stability,high uniqueness and noncontact acquisition,iris has received extensive attention.After many years of development,the accuracy of iris recognition is enough to be used in real life.The popularity of biometric technology has attracted people's attention to the problem of the leakage,loss and abuse of the biological feature template.The protection of Biometric template is a bottleneck in the development of Biometric-based authentication technology.The main techniques for preserving the privacy of biometric data are fuzzy commitment,cancelable biometrics and security multi-party computation.The first two methods have a tradeoff between the security and the performances.Iris security schemes based on Secure Multi-party Computation(SMC)can protect the privacy of the iris data,but their high computation and communication complexity are not suitable for the actual application.Because most of the iris features do not have rotation invariance,it is necessary to rotate the iris multi-times in matching stage,which increases the time of iris recognition.Li Xiang's PCET-based iris feature extraction scheme has solve the problem of iris rotation,which can reduce the matching times of iris match,but the iris feature extraction method only got 1.19% equal error rate.In this paper,we make three improvements on his method,including the selection of Gabor filter parameters,the calculation of PCET and the feature selection.The equal error rate of the improved feature extraction scheme is declined to 0.19%.A two-party security computation protocol is designed to guarantees the privacy of the client and the server during iris matching.We propose a series of optimization methods to accelerate the matching speed of iris authentication.Experimental results demonstrate that the system can effectively reduce the computational complexity and communication delay.