Bio-Discretization: Biometrics Authentication Featuring Face Data and Tokenised Random Number

Abstract

With the wonders of the Internet and the promises of the worldwide information infrastructure, a highly secure authentication system is desirable. Biometric has been deployed in this purpose as it is a unique identifier. However, it also suffers from inherent limitations and specific security threats such as biometric fabrication. To alleviate the liabilities of the biometric, a combination of token and biometric for user authentication and verification is introduced. All user data is kept in the token and human can get rid of the task of remembering passwords. The proposed framework is named as Bio- Discretization. Bio-Discretization is performed on the face image features, which is generated from Non-Negative Matrix Factorization (NMF) in the wavelet domain to produce a set of unique compact bitstring by iterated inner product between a set of pseudo random numbers and face images. Bio- Discretization possesses high data capture offset tolerance, with highly correlated bitstring for intraclass data. This approach is highly desirable in a secure environment and it outperforms the classic authentication scheme.