Voice Anti-spoofing

PartialSpoof is a dataset of partially-spoofed data to evaluate detection of partially-spoofed speech data. It has been built based on the ASVspoof 2019 LA database since the latter covers 17 types of spoofed data produced by advanced speech synthesizers, voice converters, and hybrids. The authors used the same set of bona fide data from the ASVspoof 2019 LA database but created partially spoofed audio from the ASVspoof 2019 LA data.

We introduce a new database of voice recordings with the goal of supporting research on vulnerabilities and protection of voice-controlled systems. In contrast to prior efforts, the proposed database contains genuine and replayed recordings of voice commands obtained in realistic usage scenarios and using state-of-the-art voice assistant development kits. Specifically, the database contains recordings from four systems (each with a different microphone array) in a variety of environmental conditions with different forms of background noise and relative positions between speaker and device. To the best of our knowledge, this is the first database that has been specifically designed for the protection of voice controlled systems (VCS) against various forms of replay attacks.

ASVspoof 5 is the fifth edition in a series of challenges which promote the study of speech spoofing and deepfake attacks as well as the design of detection solutions. We introduce the ASVspoof 5 database which is generated in a crowdsourced fashion from data collected in diverse acoustic conditions (cf. studio-quality data for earlier ASVspoof databases) and from ~2,000 speakers (cf. ~100 earlier). The database contains attacks generated with 32 different algorithms, also crowdsourced, and optimised to varying degrees using new surrogate detection models. Among them are attacks generated with a mix of legacy and contemporary text-to-speech synthesis and voice conversion models, in addition to adversarial attacks which are incorporated for the first time. ASVspoof 5 protocols comprise seven speaker-disjoint partitions. They include two distinct partitions for the training of different sets of attack models, two more for the development and evaluation of surrogate detection models, and then three additional partitions which comprise the ASVspoof 5 training, development and evaluation sets. An auxiliary set of data collected from an additional 30k speakers can also be used to train speaker encoders for the implementation of attack algorithms. Also described herein is an experimental validation of the new ASVspoof 5 database using a set of automatic speaker verification and spoof/deepfake baseline detectors. With the exception of protocols and tools for the generation of spoofed/deepfake speech, the resources described in this paper, already used by participants of the ASVspoof 5 challenge in 2024, are now all freely available to the community.

This dataset contains two types of audio recordings. The first set of audio recordings consists of MEMS microphone response to acoustic activities (e.g., 19 participants reading provided text in front of the Google Home Smart Assistant). The second set of audio recordings consists of MEMS microphone response to photo-acoustic activities (laser modulated--with audio recordings of 19 participants, firing at the MEMS microphone of Google Home Smart Assistant). A total of 19 students (10 male and 9 female) were enrolled for data collection. All participants were asked to read the following 5 sentences in the microphone, Hey Google, Open the garage door, Hey Google, Close the garage door, Hey Google, Turn the light on, Hey Google, Turn the light off, Hey Google, What is the weather today?. Each audio sample was injected into the microphone through a laser, and the response of the microphone was recorded. This method produced a total data set of 95 acoustic- and 95 laser-induced audio recordings.

All existing databases of spoofed speech contain attack data that is spoofed in its entirety. In practice, it is entirely plausible that successful attacks can be mounted with utterances that are only partially spoofed. By definition, partially-spoofed utterances contain a mix of both spoofed and bona fide segments, which will likely degrade the performance of countermeasures trained with entirely spoofed utterances. This hypothesis raises the obvious question: ‘Can we detect partially spoofed audio?’ This paper introduces a new database of partially-spoofed data, named PartialSpoof, to help address this question. This new database enables to investigate and compare the performance of countermeasures on both utterance- and segmental- level labels. Source: https://zenodo.org/record/4817532#.YMi9-jZKgox