mlsanityData to 14 Oct 2026

Multilingual Dataset Integration Strategies for Robust Audio Deepfake Detection: A SAFE Challenge System

arXiv 2508.20983 ↗First posted 28 Aug 2025eess.AS cs.LGcited by 0

Not on the map. Why?

Abstract (arXiv)

The SAFE Challenge evaluates synthetic speech detection across three tasks: unmodified audio, processed audio with compression artifacts, and laundered audio designed to evade detection. We systematically explore self-supervised learning (SSL) front-ends, training data compositions, and audio length configurations for robust deepfake detection. Our AASIST-based approach incorporates WavLM large frontend with RawBoost augmentation, trained on a multilingual dataset of 256,600 samples spanning 9 languages and over 70 TTS systems from CodecFake, MLAAD v5, SpoofCeleb, Famous Figures, and MAILABS. Through extensive experimentation with different SSL front-ends, three training data versions, and two audio lengths, we achieved second place in both Task 1 (unmodified audio detection) and Task 3 (laundered audio detection), demonstrating strong generalization and robustness.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Builds on 53 of its 66 references that are in the map. First posted on arXiv 28 Aug 2025.

Builds on

53 of 66 references.

References · 53

  1. 2020wav2vec 2.0: A Framework for Self-Supervised Learning of Speech Representationscited by 3,905
  2. 2021HuBERT: Self-Supervised Speech Representation Learning by Masked Prediction of Hidden Unitscited by 2,342
  3. 2021WavLM: Large-Scale Self-Supervised Pre-Training for Full Stack Speech Processingcited by 1,657
  4. 2022data2vec: A General Framework for Self-supervised Learning in Speech, Vision and Languagecited by 727
  5. 2022High Fidelity Neural Audio Compressioncited by 687
  6. 2021SUPERB: Speech processing Universal PERformance Benchmarkcited by 666
  7. 2021XLS-R: Self-supervised Cross-lingual Speech Representation Learning at Scalecited by 601
  8. 2021Grad-TTS: A Diffusion Probabilistic Model for Text-to-Speechcited by 447
  9. 2024F5-TTS: A Fairytaler that Fakes Fluent and Faithful Speech with Flow Matchingcited by 361
  10. 2023NaturalSpeech 2: Latent Diffusion Models are Natural and Zero-Shot Speech and Singing Synthesizerscited by 235
  11. 2019Mockingjay: Unsupervised Speech Representation Learning with Deep Bidirectional Transformer Encoderscited by 198
  12. 2020TERA: Self-Supervised Learning of Transformer Encoder Representation for Speechcited by 188
38 more
  1. 2019An Unsupervised Autoregressive Model for Speech Representation Learningcited by 186
  2. 2024XTTS: a Massively Multilingual Zero-Shot Text-to-Speech Modelcited by 179
  3. 2021SSAST: Self-Supervised Audio Spectrogram Transformercited by 169
  4. 2021ASVspoof 2021: accelerating progress in spoofed and deepfake speech detectioncited by 161
  5. 2023StyleTTS 2: Towards Human-Level Text-to-Speech through Style Diffusion and Adversarial Training with Large Speech Language Modelscited by 152
  6. 2024E2 TTS: Embarrassingly Easy Fully Non-Autoregressive Zero-Shot TTScited by 152
  7. 2023Matcha-TTS: A fast TTS architecture with conditional flow matchingcited by 145
  8. 2022Does Audio Deepfake Detection Generalize?cited by 135
  9. 2022ASVspoof 2021: Towards Spoofed and Deepfake Speech Detection in the Wildcited by 121
  10. 2022Automatic speaker verification spoofing and deepfake detection using wav2vec 2.0 and data augmentationcited by 117
  11. 2020Unsupervised pretraining transfers well across languagescited by 101
  12. 2024ASVspoof 5: Crowdsourced Speech Data, Deepfakes, and Adversarial Attacks at Scalecited by 91
  13. 2022MAE-AST: Masked Autoencoding Audio Spectrogram Transformercited by 82
  14. 2020DeCoAR 2.0: Deep Contextualized Acoustic Representations with Vector Quantizationcited by 77
  15. 2022SUPERB-SG: Enhanced Speech processing Universal PERformance Benchmark for Semantic and Generative Capabilitiescited by 77
  16. 2023ML-SUPERB: Multilingual Speech Universal PERformance Benchmarkcited by 71
  17. 2020Vector-Quantized Autoregressive Predictive Codingcited by 69
  18. 2021RawBoost: A Raw Data Boosting and Augmentation Method applied to Automatic Speaker Verification Anti-Spoofingcited by 67
  19. 2019Deep Contextualized Acoustic Representations For Semi-Supervised Speech Recognitioncited by 66
  20. 2021ASVspoof 2019: spoofing countermeasures for the detection of synthesized, converted and replayed speechcited by 58
  21. 2024MLAAD: The Multi-Language Audio Anti-Spoofing Datasetcited by 56
  22. 2020Non-Autoregressive Predictive Coding for Learning Speech Representations from Local Dependenciescited by 47
  23. 2021UniSpeech-SAT: Universal Speech Representation Learning with Speaker Aware Pre-Trainingcited by 45
  24. 2025Deepfake-Eval-2024: A Multi-Modal In-the-Wild Benchmark of Deepfakes Circulated in 2024cited by 45
  25. 2023FunCodec: A Fundamental, Reproducible and Integrable Open-source Toolkit for Neural Speech Codeccited by 44
  26. 2024A Large-Scale Evaluation of Speech Foundation Modelscited by 40
  27. 2024The Codecfake Dataset and Countermeasures for the Universally Detection of Deepfake Audiocited by 31
  28. 2024XLSR-Mamba: A Dual-Column Bidirectional State Space Model for Spoofing Attack Detectioncited by 30
  29. 2023Towards generalisable and calibrated synthetic speech detection with self-supervised representationscited by 23
  30. 2024SpoofCeleb: Speech Deepfake Detection and SASV In The Wildcited by 22
  31. 2024DFADD: The Diffusion and Flow-Matching Based Audio Deepfake Datasetcited by 20
  32. 2024Leveraging Mixture of Experts for Improved Speech Deepfake Detectioncited by 15
  33. 2024WavLM model ensemble for audio deepfake detectioncited by 12
  34. 2024BUT Systems and Analyses for the ASVspoof 5 Challengecited by 12
  35. 2024Exploring WavLM Back-ends for Speech Spoofing and Deepfake Detectioncited by 10
  36. 2023MiniSUPERB: Lightweight Benchmark for Self-supervised Speech Modelscited by 8
  37. 2024USTC-KXDIGIT System Description for ASVspoof5 Challengecited by 8
  38. 2021Beijing ZKJ-NPU Speaker Verification System for VoxCeleb Speaker Recognition Challenge 2021cited by 5

50 most cited of 53 All 53 (JSON)

Built on by

Not cited yet.