Research On Multi-biometric Encryption Algorithm Based On The Feature Level Fusion Of Fingerprint And Face Features

In recent years,biometric recognition technology,especially multi-biometric recognition technology,provides a convenient,reliable and secure implement method for identity authentication.However,due to the uniqueness and invariability of human biometrics,once the biometric template is leaked,it will be permanently lost and it will induce dangerous attack for users,especially for multi-biometric recognition system.In order to improve the security of biometric template and the reliability of identity authentication in multi-biometric recognition system,this thesis will focus on the research of multi-biometric encryption algorithm for multi-biometric recognition system using the most widely used fingerprint features and face features,and proposes a cancellable multi-biometric template generation algorithm based on Bloom filter and a fuzzy vault scheme based on the feature level fusion of fingerprint and face features,respectively.These algorithms can achieve reliable authentication while ensure the security of fingerprint features and face features.The main research work is as follows:(1)Firstly,this thesis analyzes all steps of fingerprint minutiae extraction algorithm,including the preprocessing of fingerprint image and the extraction of minutiae of fingerprint.Secondly,the principle of principal component analysis(PCA)is analyzed and how to extract the features of face images are studied.Finally,the fusion technology used in multi-biometric identification system is summarized,and each fusion technology is analyzed and compared in terms of the information amount,the difficulty of realization and the security of biometric templates.(2)For the security issue of multi-biometric templates in current multi-biometric recognition system,this thesis proposes a cancellable multi-biometric template generation algorithm based on Bloom filter.The algorithm uses the XOR operation in a grouped combination to fuse binary fingerprint features and binary face features into one template,and transforms the fusion template based on the irreversible characteristics of the Bloom filter,the cancellability and diversity of the fusion template by updating the random matrix.Finally,the scheme uses the traversal matching method to calculate the matching score between the templates.The experimental results show that the proposed algorithm does not need to register fingerprint image,and it can encrypt multi-biometric templates while compress the multi-biometric templates to ensure the reliability of authentication and the security of multi-biometric templates.(3)For the current issue of most of the existed biometric-based fuzzy vaults vulnerable to the correlation attack that results in the loss of key or biometric templates,this thesis proposes a fuzzy vault based on the feature level fusion of fingerprint and face features.The proposed scheme transforms the fingerprint features and face features using the discrete logarithm,and fuses the transformed fingerprint and face features into one template at the feature level based on Diffie-Hellman algorithm.Finally,the fusion template is used to construct the fuzzy vault,and it can provide the cancellability properties for the fuzzy vault by updating the random matrix.The experimental results show that the proposed scheme can solve the problem that multi-biometric fusion at the feature level easily leads to dimensionality disasters and biometric-based fuzzy vault vulnerable to the correlation attack.The proposed scheme has better reliability and improves the security of the multi-biometric templates.