Multi-Bit Allocation Preparing Voice Biometrics for Template - - PowerPoint PPT Presentation

multi bit allocation
SMART_READER_LITE
LIVE PREVIEW

Multi-Bit Allocation Preparing Voice Biometrics for Template - - PowerPoint PPT Presentation

Multi-Bit Allocation Preparing Voice Biometrics for Template Protection M. Paulini , C. Rathgeb , A. Nautsch H. Reichau , H. Reininger , C. Busch atip Advanced Technologies for Information Processing GmbH


slide-1
SLIDE 1

Multi-Bit Allocation

Preparing Voice Biometrics for Template Protection

  • M. Paulini⋆, C. Rathgeb†, A. Nautsch†
  • H. Reichau⋆, H. Reininger⋆, C. Busch†

⋆ atip – Advanced Technologies for Information Processing GmbH †Hochschule Darmstadt, CRISP, CASED, da/sec Security Research Group

Odyssey’16, Bilbao, 22.06.2016

Paulini et al. Multi-Bit Allocation / Odyssey’16, Bilbao, 22.06.2016 1/3

slide-2
SLIDE 2

Motivation Biometric Template Protection

◮ Biometric characteristics are NOT secret in an age of

increased surveillance

◮ The link between individuals and their biometric

(physiological) characteristics is strong and permanent Biometric reference data (templates) MUST be protected in order to safeguard individuals’ privacy and biometric systems’ security! Biometric Template Protection:

◮ Biometric cryptosystems ◮ Cancelable biometrics

Properties (ISO/IEC IS 24745):

◮ Irreversibility ◮ Unlinkability

Paulini et al. Multi-Bit Allocation / Odyssey’16, Bilbao, 22.06.2016 2/3

slide-3
SLIDE 3

Contribution Multi-Bit Allocation

Tackled issue:

◮ Biometric cryptosystems operate on a certain form of input,

e.g. binary features or point-sets (fuzzy commitment/vault)

◮ Incompatibility issues arise when the type of features does

NOT match the acceptable input type!

1000110 0110101 1011010 Feature Extraction

Feature-T ype Transformation

PE

Multi-Bit Allocation

T emplate Protection Voice Sample

Feature-type transformations are required! Proposed system:

◮ Ordered-to-ordered real-to-binary feature-type transformation,

Multi-Bit Allocation

◮ Detection of discriminative features, Relevant Projections

Paulini et al. Multi-Bit Allocation / Odyssey’16, Bilbao, 22.06.2016 3/3