Biometric template protection (BTP) is an open problem for biometric identity management systems. Cancellable biometrics is commonly designed to protect biometric templates with two input factors i.e., biometrics and a token used in template replacement. However, the token is often required to be kept secretly; otherwise, the protected template could be vulnerable to several security attacks and breaches of privacy. In this paper, we propose a tokenless cancellable biometrics scheme called Multimodal Extended Feature Vector (M∙EFV) Hashing that employs an improved XOR encryption/decryption notion to operate on the transformation key. We stress on multimodal biometrics where the real-valued face and fingerprint vectors are fused and embedded into a binarized cancellable template. Specifically, M∙EFV hashing consists of three stages of transformation: 1) normalization and biometric fusion; 2) randomization and binarization; and 3) cancellable template generation. To evaluate the proposed scheme, several benchmarking datasets, i.e., FVC2002, FVC2004 for fingerprint and LFW for face are used in experiments. The verification performance is validated by employing the FVC matching protocol. Various attacks are simulated and analysed in the worst-case scenario. Lastly, unlinkability and revocability properties are examined experimentally.
Bibliographical noteFunding Information:
Zhe Jin obtained his BIT (Hons) in Software Engineering, MSc (I.T.) from Multimedia University, Malaysia in 2007 and 2011 respectively, and Ph.D. in Engineering from University Tunku Abdul Rahman Malaysia in 2016. Currently, he is a Senior Lecturer at School of Information Technology, MONASH University, Malaysia campus. His research interests include Biometric Security, Computer Vision, and Machine Learning. He has published more than 40 refereed journals, conference articles, including IEEE Trans. IFS, SMC-S, DSC, Pattern Recognition. He was awarded Marie Skłodowska-Curie Research Exchange Fellowship and visited the University of Salzburg, Austria, and the University of Sassari, Italy respectively as a visiting scholar under the EU Project IDENTITY 690907.
This research was partly supported by the Ministry of Higher Education (MOHE) Malaysia through the Fundamental Research Grant Scheme (FRGS) (FRGS/1/2018/ICT02/MUSM/03/3). We also gratefully acknowledge the support of NVIDIA Corporation with the GPU grant donation of the Titan Xp GPU used for this research.
All Science Journal Classification (ASJC) codes
- Computer Science(all)