Datasets:
license: cc-by-nc-4.0
language:
- en
pretty_name: Deepfake and Spoof Detection meets Neural Audio Codecs
dataset_info:
features:
- name: utt_id
dtype: string
- name: origin_ds
dtype: string
- name: attacker
dtype: string
- name: label
dtype: string
- name: speaker
dtype: string
- name: audio
dtype: audio
splits:
- name: main
num_bytes: 23569231227.24
num_examples: 63473
download_size: 19759387664
dataset_size: 23569231227.24
configs:
- config_name: default
data_files:
- split: main
path: data/main-*
tags:
- deepfake
- spoof
- detection
In this spoof detection dataset, the bonafide speech is resynthesized using various popular neural audio codecs, which are used for compression and low-bandwidth transmission of speech signals. The spoofed speech samples we provide are generated with a selection of popular and well performing language model based speech synthesis methods, which utilize the same codecs as the bonafide audios to obtain discretized speech tokens. This takes the artifacts of the codecs out of the equation and lets a deepfake detection model trained on this data focus purely on higher-level patterns to differentiate the genuine human samples from the faked speech samples.
This dataset can be seen as a very challenging extension to the ASVspoof 5 dataset [1, 2] that aligns the field more closely with upcoming challenges in the real world. For compatibility, we follow the design of the data and the dev and test splits with 50% of speakers overlapping the train set and 50% being completely unseen. A publication associated with this dataset is currently under review.
The huggingface version of this dataset contains all the generated audio and the seed bonafide audio used to generate them. To reproduce our results or compare on the exact same splits, please use the version we provide at https://zenodo.org/records/17225924 which includes additionally the bonafide audio used in the dev and test splits.
| Attacker | Codec | Parameters | Applied to Bonafide | Link |
|---|---|---|---|---|
| Llasa 8B | XCodec2 | 8B | ✅ | https://huggingface.co/HKUSTAudio/Llasa-8B |
| MARS5 | EnCodec (Vocos decoder) | 1B | ✅ | https://huggingface.co/CAMB-AI/MARS5-TTS |
| CSM | Mimi | 1B | ✅ | https://huggingface.co/sesame/csm-1b |
| OpenAudio S1-mini | DescriptAudioCodec | 0.5B | ✅ | https://huggingface.co/fishaudio/openaudio-s1-mini |
| Chatterbox | S3Tokenizer | 0.5B | ❌ | https://huggingface.co/ResembleAI/chatterbox |
| CosyVoice2 | S3Tokenizer | 0.5B | ❌ | https://huggingface.co/FunAudioLLM/CosyVoice2-0.5B |
[1] 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, and Vishwanath Singh. 2024. ASVspoof 5: Design, Collection and Validation of Resources for Spoofing, Deepfake, and Adversarial Attack Detection Using Crowdsourced Speech. In Computer Speech & Language, 2026, 95. Jg., S. 101825. https://www.sciencedirect.com/science/article/pii/S0885230825000506
[2] 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, and Junichi Yamagishi. 2024. ASVspoof 5: Crowdsourced speech data, deepfakes, and adversarial attacks at scale. In ASVspoof Workshop 2024, 2024. 1--8. https://doi.org/10.21437/ASVspoof.2024-1