Long Sequence Biohashing Speech Authentication Based on Biometric Fusion and Modified Logistic Measurement Matrix

Based on the speech content authentication algorithm, the biometric of the speech are directly constructed hash and stored it in the cloud, which is easy to cause the leakage of the biometric. At the same time, the short hash sequence generated by the biometric will lead to the lack of authentication accuracy. Aiming at the security of speech biometrics and the discrimination of hash sequences in the process of authentication, this paper proposes a long sequence BioHashing speech authentication algorithm based on biometric fusion and modified logistic measurement matrix. Firstly, the pre-processed speech extracts GFCC biometric and calculates its Euclidean distance in frequency domain and TEOCC biometric is calculated in time domain, and time-frequency biometrics are fused. Then, the modified logistic measurement matrix Schmidt controlled by the key is orthogonalized to construct the orthogonal set matrix. Finally, the biosafety template is constructed by inner product of time-frequency fused biometrics and the orthogonal set matrix and a long sequence BioHashing is generated by the biosafety template of the binarization. In authentication, the Hamming distance is used to match BioHashing. The experimental results show that when the BioHashing length is 798bits, the long sequence can effectively improve the discrimination and effectively balance the robustness and discrimination of the algorithm. At the same time, measurement matrix controlled by the key can effectively reduce the risk of biometric leakage and improve the security of the biometric in the authentication process.