Neural Codec Source Tracing: Toward Comprehensive Attribution in Open-Set Condition
Abstract
Current research in audio deepfake detection is gradually transitioning from binary classification to multi-class tasks, referred as audio deepfake source tracing task. However, existing studies on source tracing consider only closed-set scenarios and have not considered the challenges posed by open-set conditions. In this paper, we define the Neural Codec Source Tracing (NCST) task, which is capable of performing open-set neural codec classification and interpretable ALM detection. Specifically, we constructed the ST-Codecfake dataset for the NCST task, which includes bilingual audio samples generated by 11 state-of-the-art neural codec methods and ALM-based out-ofdistribution (OOD) test samples. Furthermore, we establish a comprehensive source tracing benchmark to assess NCST models in open-set conditions. The experimental results reveal that although the NCST models perform well in in-distribution (ID) classification and OOD detection, they lack robustness in classifying unseen real audio. The ST-codecfake dataset and code are available.
Keywords
Cite
@article{arxiv.2501.06514,
title = {Neural Codec Source Tracing: Toward Comprehensive Attribution in Open-Set Condition},
author = {Yuankun Xie and Xiaopeng Wang and Zhiyong Wang and Ruibo Fu and Zhengqi Wen and Songjun Cao and Long Ma and Chenxing Li and Haonnan Cheng and Long Ye},
journal= {arXiv preprint arXiv:2501.06514},
year = {2025}
}