Welcome to ReplayDF!

This dataset demonstrates how replay attacks can significantly degrade the performance of state-of-the-art audio deepfake detection systems. It contains re-recorded samples derived from M-AILABS and MLAAD using 109 unique speaker-microphone combinations, covering six languages and four TTS models under diverse acoustic conditions. ReplayDF is released for non-commercial research to support the study and development of more robust deepfake detection methods.

This dataset makes use of the MLAAD and M-AILABS dataset. MLAAD provides only the synthetic audio, while M-AILABS provides the real audio.

Downloads:

• Dataset download via Huggingface Datasets.
• Paper
• Creative Commons CC BY NC 4.0

How to cite:

@article{muller2025replaydf,
  title = {Replay Attacks Against Audio Deepfake Detection},
  author = {Nicolas Müller and Piotr Kawa and Wei-Herng Choong and Adriana Stan and Aditya Tirumala Bukkapatnam and Karla Pizzi and Alexander Wagner and Philip Sperl},
  journal={Interspeech 2025},
  year = {2025},
}