Research On Key Technologies Of Privacy-Preserving For Biometric Recognition Services

The rapid development of image recognition and deep learning technology has greatly improved the availability of biometric recognition technology,and the development of mobile Internet and the popularization of intelligent mobile terminal devices have also provided a good environment for the deployment of biometric recognition services.The identity authentication and identification services based on biometric recognition have been widely used in many fields,which provides great convenience for people’s lives.However,the large-scale application of biometric recognition technology also brings great risk of disclosure of users’ biometric privacy data.To address this problem,academia has proposed a variety of privacy-preserving schemes for biometric recognition services based on technologies such as homomorphic encryption and secure multi-party computing.However,most of these existing schemes have shortcomings in terms of efficiency.Therefore,how to ensure the efficiency of biometric recognition services on the premise of ensuring the security of user’s biometric data is a current challenge.This dissertation studies the privacy protection issues in biometric recognition services,which mainly focuses on two classic biometric recognition service scenarios,namely,biometric authentication and biometric identification.By analyzing key issues such as privacy protection requirements and efficiency bottlenecks in specific application scenarios,a privacy protection scheme for specific application scenarios is constructed based on cryptography,data structure and other technologies.At the same time,the correctness,security and efficiency of the proposed scheme are theoretically analyzed,and a prototype system is constructed to verify the feasibility and efficiency of the proposed scheme on the real biometric data set.Specifically,the main contribution of this dissertation includes the following three aspects.(1)Aiming at the privacy-preserving problem in the biometric authentication scenario,a privacy-preserving biometric authentication scheme for smart home scenarios is proposed based on matrix encryption technology.The user’s biometric registration and authentication templates are encrypted through matrix encryption technology to ensure the security of biometric data in the storage process.The similarity matching between user biometric templates is computed based on their ciphertexts,which ensures the security of biometric data in the authentication process.Security analysis shows that this scheme can ensure the security of user biometric data during the execution of the scheme.Simulation experiments show that the proposed scheme has certain advantages in execution efficiency compared with similar schemes.(2)Aiming at the privacy protection problem in the biometric identification scenario,an efficient and privacy-preserving biometric identification scheme is implemented based on the symmetric homomorphic encryption scheme and the M-tree data structure,which can ensure the security of the service provider’s biometric data set and user’s biometric information during the biometric identification process.By introducing the M-tree data structure,the computational costs of the cloud server while searching for target template in the biometric data set is significantly reduced,and the execution efficiency of the biometric identification scheme is improved.The security analysis proves that the proposed scheme can ensure the security of biometric information in every stage of the identification service,and the simulation experiment proves that the proposed scheme has obvious advantages in terms of computational and communication costs compared with the identification scheme without the introduction of optimization strategies.(3)Aiming at the privacy-preserving problem in biometric identification scenario,an efficient and privacy-preserving biometric identification scheme is proposed based on symmetric homomorphic encryption scheme and Fiting-tree data structure.On the premise of ensuring the security of biometric data,a data structure which can achieve fast query on biometric data sets is designed by combining Fiting-tree and i Distance index,which reduces the computational and communication overhead of cloud servers in the search process.The security of the scheme is analyzed from the perspective of the data that each participant can access in the biometric identification process.Experiments on simulated datasets demonstrate that the proposed scheme has obvious advantages over M-tree-based recognition schemes in terms of computational and communication overhead.At the same time,the recognition accuracy of the proposed scheme is verified through simulation experiments on real face datasets.