Speech biometric mapping for key binding cryptosystem
Keerati Inthavisas, Daniel Lopresti · Proceedings of SPIE, the International Society for Optical Engineering/Proceedings of SPIE · 2011
We propose a new scheme to transform speech biometric measurements (feature vector) to a binary string which can be combined with a pseudo-random key for a cryptographic purpose. We utilize Dynamic Time Warping (DTW) in our scheme. The challenge of using DTW in a cryptosystem is that a template must be useful to create a warping function, while it must not be usable for an attacker to derive the cryptographic key. In this work, we propose a hardened template to address these problems. We evaluate our scheme with two speech datasets and compare with DTW, VQ, and GMM speaker verifications. The experimental results show that the performance of the proposed scheme outperforms VQ and GMM. It is slightly degraded when compared to the DTW speaker verification. The EERs against attackers utilizing the hardened template are 0% both datasets.