Abstract
In this work, we introduce the task of singing voice deep fake source attribution (SVDSA). We hypothesize that multi modal foundation models (MMFMs) such as ImageBind, Lan guageBind will be most effective for SVDSA as they are bet ter equipped for capturing subtle source-specific characteris tics—such as unique timbre, pitch manipulation, or synthesis artifacts of each singing voice deepfake source due to their cross modality pre-training. Our experiments with MMFMs, speech foundation models and music foundation models verify the hy pothesis that MMFMs are the most effective for SVDSA. Fur thermore, inspired from related research, we also explore fusion of foundation models (FMs) for improved SVDSA. To this end, we propose a novel framework, COFFE which employs Cher noff Distance as novel loss function for effective fusion of FMs. Through COFFE with the symphony of MMFMs, we attain the topmost performance in comparison to all the individual FMs and baseline fusion methods. Index Terms: Source Attribution, Singing Voice Deepfake, Deepfake Detection

If you find this work useful, please consider citing us:

@misc{phukan2025sourceattributionsingingvoice,
    title={Towards Source Attribution of Singing Voice Deepfake with Multimodal Foundation Models}, 
    author={Orchid Chetia Phukan and Girish and Mohd Mujtaba Akhtar and Swarup Ranjan Behera and Priyabrata Mallick and Pailla Balakrishna Reddy and Arun Balaji Buduru and Rajesh Sharma},
    year={2025},
    eprint={2506.03364},
    archivePrefix={arXiv},
    primaryClass={eess.AS},
    url={https://arxiv.org/abs/2506.03364}, 
}