ASVspoof 5: Design, Collection and Validation of Resources for Spoofing, Deepfake, and Adversarial Attack Detection Using Crowdsourced Speech

View on arXiv ← Back to list

Authors: Xin Wang, Héctor Delgado, Hemlata Tak, Jee-weon Jung, Hye-jin Shim, Massimiliano Todisco, Ivan Kukanov, Xuechen Liu, Md Sahidullah, Tomi Kinnunen, Nicholas Evans, Kong Aik Lee, Junichi Yamagishi, Myeonghun Jeong, Ge Zhu, Yongyi Zang, You Zhang, Soumi Maiti, Florian Lux, Nicolas Müller, Wangyou Zhang, Chengzhe Sun, Shuwei Hou, Siwei Lyu, Sébastien Le Maguer, Cheng Gong, Hanjie Guo, Liping Chen, Vishwanath Singh

Published: 2025-02-13 00:15:54+00:00

Comment: Database link: https://zenodo.org/records/14498691, Database mirror link: https://huggingface.co/datasets/jungjee/asvspoof5, ASVspoof 5 Challenge Workshop Proceeding: https://www.isca-archive.org/asvspoof_2024/index.html

AI Summary

ASVspoof 5 introduces a new challenge and a comprehensive crowdsourced database designed for evaluating speech spoofing, deepfake, and adversarial attack detection solutions. The database features speech from approximately 2,000 speakers, incorporating diverse acoustic conditions and attacks generated by 32 different algorithms, including legacy, contemporary TTS/VC, and adversarial methods. The paper details the database design, collection, and experimental validation using baseline detectors, making the resources freely available to the research community.

Abstract

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.

Key findings

The validation experiments reveal that both legacy (e.g., unit-selection TTS) and contemporary deepfake technologies, along with adversarial attacks, pose a significant threat to automatic speaker verification and spoofing detection systems. Encoding and compression conditions further exacerbate the challenge, leading to notable performance degradation for baseline detectors. The design successfully reduced shortcut artifacts, ensuring that detection systems learn semantically relevant cues.

Approach

The paper's approach is to provide a standardized, robust framework for evaluating audio deepfake detection systems. This involves designing and collecting a large-scale crowdsourced database with diverse bona fide and spoofed speech, generated by 32 different TTS, VC, and adversarial attack algorithms under varied acoustic conditions. It also establishes evaluation protocols, including steps to reduce shortcut artifacts, and validates the database using baseline ASV, CM, and SASV systems.

Datasets

ASVspoof 5 database, MLS English database, LibriVox, Common Voice (ver. 11.0), VoxCeleb1, VoxCeleb2, MUSAN, RIR, VoxPopuli.

Model(s)

ECAPA-TDNN, RawNet2, AASIST, LCNNs, MFA-Conformer, wav2vec 2.0, CAM++.

Author countries

Japan, Spain, USA, Republic of Korea, France, Singapore, India, Finland, China, Germany.

← Previous