The 2nd Automatic Speaker Verification Spoofing and Countermeasures Challenge (ASVspoof 2017) Database, Version 2

Abstract

This is a database used for the Second Automatic Speaker Verification Spoofing and Countermeasuers Challenge, for short, ASVspoof 2017 (http://www.asvspoof.org) organized by Tomi Kinnunen, Md Sahidullah, Héctor Delgado, Massimiliano Todisco, Nicholas Evans, Junichi Yamagishi, Kong Aik Lee in 2017. The ASVspoof challenge aims to encourage further progress through (i) the collection and distribution of a standard dataset with varying spoofing attacks implemented with multiple, diverse algorithms and (ii) a series of competitive evaluations for automatic speaker verification. The ASVspoof 2017 challenge follows on from two special sessions on spoofing and countermeasures for automatic speaker verification held during INTERSPEECH 2013 and 2015. While the first edition in 2013 was targeted mainly at increasing awareness of the spoofing problem, the 2015 edition included a first challenge on the topic, with commonly defined evaluation data, metrics and protocols. The task in ASVspoof 2015 was to discriminate genuine human speech from speech produced using text-to-speech (TTS) and voice conversion (VC) attacks. The challenge was drawn upon state-of-the-art TTS and VC attacks data prepared for the “SAS” corpus by TTS and VC researchers. The primary technical goal of ASVspoof 2017 is to assess spoofing attack detection accuracy with ‘out in the wild’ conditions, thereby advancing research towards generalized spoofing countermeasure, in particular to detect replay. In addition, ASVspoof 2017 attempts to better interlink the research efforts from spoofing and text-dependent ASV communities. To this end, ASVspoof 2017 makes an extensive use of the recent text-dependent RedDots corpus, as well as a replayed version of the same data. The ASVspoof 2017 database contains large amount of speech data collected from 179 replay sessions in 61 unique replay configurations. Number of speakers is 42. A replay configuration means a unique combination of room, replay device and recording device, while a session refers to a set of source files, which share the same replay configuration. This is version 2 of the ASVspoof 2017 database. Please see README and ChangeLog for more details.

This is a database used for the Second Automatic Speaker Verification Spoofing and Countermeasuers Challenge, for short, ASVspoof 2017 (http://www.asvspoof.org) organized by Tomi Kinnunen, Md Sahidullah, Héctor Delgado, Massimiliano Todisco, Nicholas Evans, Junichi Yamagishi, Kong Aik Lee in 2017. The ASVspoof challenge aims to encourage further progress through (i) the collection and distribution of a standard dataset with varying spoofing attacks implemented with multiple, diverse algorithms and (ii) a series of competitive evaluations for automatic speaker verification. The ASVspoof 2017 challenge follows on from two special sessions on spoofing and countermeasures for automatic speaker verification held during INTERSPEECH 2013 and 2015. While the first edition in 2013 was targeted mainly at increasing awareness of the spoofing problem, the 2015 edition included a first challenge on the topic, with commonly defined evaluation data, metrics and protocols. The task in ASVspoof 2015 was to discriminate genuine human speech from speech produced using text-to-speech (TTS) and voice conversion (VC) attacks. The challenge was drawn upon state-of-the-art TTS and VC attacks data prepared for the “SAS” corpus by TTS and VC researchers. The primary technical goal of ASVspoof 2017 is to assess spoofing attack detection accuracy with ‘out in the wild’ conditions, thereby advancing research towards generalized spoofing countermeasure, in particular to detect replay. In addition, ASVspoof 2017 attempts to better interlink the research efforts from spoofing and text-dependent ASV communities. To this end, ASVspoof 2017 makes an extensive use of the recent text-dependent RedDots corpus, as well as a replayed version of the same data. The ASVspoof 2017 database contains large amount of speech data collected from 179 replay sessions in 61 unique replay configurations. Number of speakers is 42. A replay configuration means a unique combination of room, replay device and recording device, while a session refers to a set of source files, which share the same replay configuration. This is version 2 of the ASVspoof 2017 database. Please see README and ChangeLog for more details.