FSD: An Initial Chinese Dataset for Fake Song Detection

TL;DR

This paper introduces the first Chinese dataset for fake song detection, demonstrating that models trained specifically on songs outperform speech-trained models in identifying deepfake songs.

Contribution

The paper creates a novel Chinese Fake Song Detection dataset and shows that song-trained models significantly improve deepfake detection accuracy over speech-trained models.

Findings

Song-trained ADD models reduce error rates by 38.58% compared to speech-trained models.

Existing speech-trained models are ineffective for song deepfake detection.

The dataset includes fake songs generated by five advanced synthesis methods.

Abstract

Singing voice synthesis and singing voice conversion have significantly advanced, revolutionizing musical experiences. However, the rise of "Deepfake Songs" generated by these technologies raises concerns about authenticity. Unlike Audio DeepFake Detection (ADD), the field of song deepfake detection lacks specialized datasets or methods for song authenticity verification. In this paper, we initially construct a Chinese Fake Song Detection (FSD) dataset to investigate the field of song deepfake detection. The fake songs in the FSD dataset are generated by five state-of-the-art singing voice synthesis and singing voice conversion methods. Our initial experiments on FSD revealed the ineffectiveness of existing speech-trained ADD models for the task of song deepFake detection. Thus, we employ the FSD dataset for the training of ADD models. We subsequently evaluate these models under two scenarios: one with the original songs and another with separated vocal tracks. Experiment results show that song-trained ADD models exhibit a 38.58% reduction in average equal error rate compared to speech-trained ADD models on the FSD test set.