Optimized helper data scheme for biometric verification under zero leakage constraint

In biometric verication, special measures are needed to prevent that a dishon- est verier can steal privacy-sensitive information about the prover from the template database. We introduce an improved version of the zero leakage quan- tization scheme, which optimizes detection performance in terms of the false rejection ratio. Our scheme ensures zero leakage, that is, zero mutual informa- tion between auxiliary verication data and the protected enrolled secret and guarantees a uniformly distributed secret. Moreover, our solution replaces the helper data in the template database by a user specic threshold and eliminates the pre-distortion in the verication phase, which allows very rapid verication algorithms. Although the false rejection rate was only reduced by 20% for well correlated biometric features, it can be shown that this particular scheme achieves optimal detection under our requirements.