Cancelable multi-biometrics: Mixing iris-codes based on adaptive bloom filters

• Published in: Computers & security Vol. 42; pp. 1 - 12
• Main Authors: Rathgeb, C., Busch, C.
• Format: Journal Article
• Language: English
• Published: Amsterdam Elsevier Ltd 01.05.2014; Elsevier; Elsevier Sequoia S.A
• Subjects: Applied sciences; Biometrics; Cancelable biometrics; Compressing; Computer information security; Computer science; control theory; systems; Computer systems and distributed systems. User interface; Data processing. List processing. Character string processing; Detection, estimation, filtering, equalization, prediction; Errors; Exact sciences and technology; Eyes; Eyes & eyesight; Feature extraction; Fourier transforms; Information, signal and communications theory; Memory and file management (including protection and security); Memory organisation. Data processing; Multi-biometrics; Security; Signal and communications theory; Signal, noise; Software; Studies; Telecommunications and information theory; Template protection; Transforms; Biometrics; Security; Template protection; Multi-biometrics; Cancelable biometrics; Private life; Mixing; Error rate; Anonymity; Irreversible process; Bloom filter; Adaptive filter; Adaptive coding; Dimension reduction; Digital filter; Selection criterion; Data fusion; Feature extraction; Computer security; Pattern extraction
• ISSN: 0167-4048; 1872-6208
• DOI: 10.1016/j.cose.2013.12.005

In this work adaptive Bloom filter-based transforms are applied in order to mix binary iris biometric templates at feature level, where iris-codes are obtained from both eyes of a single subject. The irreversible mixing transform, which generates alignment-free templates, obscures information present in different iris-codes. In addition, the transform is parameterized in order to achieve unlinkability, implementing cancelable multi-biometrics. Experiments which are carried out on the IITD Iris Database version 1.0 confirm the soundness of the proposed approach, (1) maintaining biometric performance at equal error rates below 0.5% for different feature extraction methods and fusion scenarios and (2) achieving a compression of mixed templates down to 10% of original size.