Ecole d'ingénieur et centre de recherche en Sciences du numérique

A one-class classification approach to generalised speaker verification spoofing countermeasures using local binary patterns

Alegre, Federico; Amehraye, Asmaa; Evans, Nicholas

BTAS 2013, 6th IEEE International Conference on Biometrics: Theory, Applications and Systems, Washington, September 29-October 2, 2013, Washington DC, USA

The vulnerability of automatic speaker verification systems to spoofing is now well accepted. While recent work has shown the potential to develop countermeasures capable of detecting spoofed speech signals, existing solutions typically function well only for specific attacks on which they are optimised. Since the exact nature of spoofing attacks can never be known in practice, there is thus a need for generalised countermeasures which can detect previously unseen spoofing attacks. This paper presents a novel countermeasure based on the analysis of speech signals using local binary patterns followed by a one-class classification approach. The new countermeasure captures differences in the spectro-temporal texture of genuine and spoofed speech, but relies only on a model of the former. We report experiments with three different approaches to spoofing and with a state-of-the-art i-vector speaker verification system which uses probabilistic linear discriminant analysis for intersession compensation. While a support vector machine classifier is tuned with examples of converted voice, it delivers reliable detection of spoofing attacks using synthesized speech and artificial signals, attacks for which it is not optimised.

Document Doi Hal Bibtex

Titre:A one-class classification approach to generalised speaker verification spoofing countermeasures using local binary patterns
Type:Conférence
Langue:English
Ville:Washington
Pays:ÉTATS-UNIS
Date:
Département:Sécurité numérique
Eurecom ref:4093
Copyright: © 2013 IEEE. Personal use of this material is permitted. However, permission to reprint/republish this material for advertising or promotional purposes or for creating new collective works for resale or redistribution to servers or lists, or to reuse any copyrighted component of this work in other works must be obtained from the IEEE.
Bibtex: @inproceedings{EURECOM+4093, doi = {http://dx.doi.org/10.1109/BTAS.2013.6712706}, year = {2013}, title = {{A} one-class classification approach to generalised speaker verification spoofing countermeasures using local binary patterns}, author = {{A}legre, {F}ederico and {A}mehraye, {A}smaa and {E}vans, {N}icholas}, booktitle = {{BTAS} 2013, 6th {IEEE} {I}nternational {C}onference on {B}iometrics: {T}heory, {A}pplications and {S}ystems, {W}ashington, {S}eptember 29-{O}ctober 2, 2013, {W}ashington {DC}, {USA}}, address = {{W}ashington, {\'{E}}{TATS}-{UNIS}}, month = {09}, url = {http://www.eurecom.fr/publication/4093} }
Voir aussi: