Towards cancelable multi-biometrics based on bloom filters: a case study on feature level fusion of face and iris

In this work we propose a generic framework for generating an irreversible representation of multiple biometric templates based on adaptive Bloom filters. The presented technique enables a feature level fusion of different biometrics (face and iris) to a single protected template, improving privacy protection compared to the corresponding systems based on a single biometric trait. At the same time, a significant gain in biometric performance is achieved, confirming the sound- ness of the proposed technique.