Security Evaluation on Speech Biometric Authentication System
Keyword(s):
A recent work utilized a transformation function to protect a DTW template. Unfortunately, a matching template was not protected properly. In this paper, we first show that an adversary can exploit the matching template to gain access to the system. Then, we introduce our scheme to address this problem. For this scheme, a hardened template is utilized to protect the DTW template. For the matching template, it is protected by a cryptographic framework. We evaluate the system with a public database: the MIT mobile device speaker verification corpus. The experimental results show that our scheme outperforms the other approaches.