This paper describes the scientific study of vulnerability of automatic speaker verification to various vary of spoofing attacks. We start with radical analysis of the spoofing effects of five speech syntheses and eight voice conversion system and also the vulnerability of three speaker verification system under those attacks. Then we introduce variety of countermeasures to prevent the spoofing attacks from the each noted and unknown attackers. Known attacker’s square measure spoofing system those output was used to train the countermeasures, whereas associate unknown assailant is a spoofing system whose output offers to the countermeasures during coaching. Finally we describe benchmark automatic system against human performance on each speaker verification and spoofing deduction task. The main objective of the proposed method is to protect from the spoofing attack for text independent speaker verification.
[1]
Mikhail Khitrov,et al.
Talking passwords: voice biometrics for data access and security
,
2013
.
[2]
Zhizheng Wu,et al.
Human vs machine spoofing detection on wideband and narrowband data
,
2015,
INTERSPEECH.
[3]
Brett Beranek.
Voice biometrics: success stories, success factors and what's next
,
2013
.
[4]
Paul Golden.
Voice biometrics – the Asia Pacific experience
,
2012
.
[5]
Tomoki Toda,et al.
SAS: A speaker verification spoofing database containing diverse attacks
,
2015,
2015 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP).