English

ERF-BA-TFD+: A Multimodal Model for Audio-Visual Deepfake Detection

Artificial Intelligence 2025-12-04 v2 Sound

Abstract

Deepfake detection is a critical task in identifying manipulated multimedia content. In real-world scenarios, deepfake content can manifest across multiple modalities, including audio and video. To address this challenge, we present ERF-BA-TFD+, a novel multimodal deepfake detection model that combines enhanced receptive field (ERF) and audio-visual fusion. Our model processes both audio and video features simultaneously, leveraging their complementary information to improve detection accuracy and robustness. The key innovation of ERF-BA-TFD+ lies in its ability to model long-range dependencies within the audio-visual input, allowing it to better capture subtle discrepancies between real and fake content. In our experiments, we evaluate ERF-BA-TFD+ on the DDL-AV dataset, which consists of both segmented and full-length video clips. Unlike previous benchmarks, which focused primarily on isolated segments, the DDL-AV dataset allows us to assess the model's performance in a more comprehensive and realistic setting. Our method achieves state-of-the-art results on this dataset, outperforming existing techniques in terms of both accuracy and processing speed. The ERF-BA-TFD+ model demonstrated its effectiveness in the "Workshop on Deepfake Detection, Localization, and Interpretability," Track 2: Audio-Visual Detection and Localization (DDL-AV), and won first place in this competition.

Keywords

Cite

@article{arxiv.2508.17282,
  title  = {ERF-BA-TFD+: A Multimodal Model for Audio-Visual Deepfake Detection},
  author = {Xin Zhang and Jiaming Chu and Jian Zhao and Yuchu Jiang and Xu Yang and Lei Jin and Chi Zhang and Xuelong Li},
  journal= {arXiv preprint arXiv:2508.17282},
  year   = {2025}
}

Comments

The paper is withdrawn after discovering a flaw in the theoretical derivation presented in Section Method. The incorrect step leads to conclusions that are not supported by the corrected derivation. We plan to reconstruct the argument and will release an updated version once the issue is fully resolved

R2 v1 2026-07-01T05:03:20.652Z