The dataset viewer is not available because its heuristics could not detect any supported data files. You can try uploading some data files, or configuring the data files location manually.
The ADD-C test dataset is proposed to assess the performance of ADD systems in real-world communication scenarios.
ADD-C includes six evaluation conditions, a clean condition (C0), and five real-world communication degraded conditions (C1-C5). Each real-world condition includes simulating audio codec compression using six widely used speech codecs in VoLTE and VoIP communication systems, as well as simulating channel transmission effects under five different Packet Loss Rates (PLRs).
ADD-C comprises:
-124000 audio files (1) C0: 2000 real, 2000 fake, without any real-world communication degradations (clean and high fidelity audio samples) (2) C1: 12000 real, 12000 fake, with 6 codec compression under packet loss rate = 0% (3) C2: 12000 real, 12000 fake, with 6 codec compression under packet loss rate = 1% (4) C3: 12000 real, 12000 fake, with 6 codec compression under packet loss rate = 5% (5) C4: 12000 real, 12000 fake, with 6 codec compression under packet loss rate = 10% (6) C5: 12000 real, 12000 fake, with 6 codec compression under packet loss rate = 20%
-included 36 spoofing algorithms
The dataset is supposed to assess the robustness of ADD systems under various communication conditions.
Further information, please refer to our paper: "Benchmarking Audio Deepfake Detection Robustness in Real-world Communication Scenarios" (https://eusipco2025.org/wp-content/uploads/pdfs/0000566.pdf)
How to cite: @article{haohan, title={Benchmarking Audio Deepfake Detection Robustness in Real-world Communication Scenarios}, author={Shi, Haohan and Shi, Xiyu and Dogan, Safak and Alzubi, Saif and Huang, Tianjin and Zhang, Yunxiao}, journal={in Proc. of the 33rd European Signal Processing Conference (EUSIPCO)}, pages={566--570}, year={2025} }
- Downloads last month
- 8