Beyond Silence: Bias Analysis through Loss and Asymmetric Approach in Audio Anti-Spoofing
Not on the map. Why?
Abstract (arXiv)
Current trends in audio anti-spoofing detection research strive to improve models' ability to generalize across unseen attacks by learning to identify a variety of spoofing artifacts. This emphasis has primarily focused on the spoof class. Recently, several studies have noted that the distribution of silence differs between the two classes, which can serve as a shortcut. In this paper, we extend class-wise interpretations beyond silence. We employ loss analysis and asymmetric methodologies to move away from traditional attack-focused and result-oriented evaluations towards a deeper examination of model behaviors. Our investigations highlight the significant differences in training dynamics between the two classes, emphasizing the need for future research to focus on robust modeling of the bonafide class.
Key facts. Cited by 2 papers in the map (data through 7 Oct 2026). Builds on 26 of its 43 references that are in the map. First posted on arXiv 25 Jun 2024.
Builds on
26 of 43 references.
References · 26
- 2020Shortcut Learning in Deep Neural Networkscited by 1,928
- 2020Sharpness-Aware Minimization for Efficiently Improving Generalizationcited by 1,404
- 2018Generalized Cross Entropy Loss for Training Deep Neural Networks with Noisy Labelscited by 994
- 2020CurricularFace: Adaptive Curriculum Learning Loss for Deep Face Recognitioncited by 320
- 2021ASAM: Adaptive Sharpness-Aware Minimization for Scale-Invariant Learning of Deep Neural Networkscited by 293
- 2021AASIST: Audio Anti-Spoofing using Integrated Spectro-Temporal Graph Attention Networkscited by 173
- 2020End-to-end anti-spoofing with RawNet2cited by 160
- 2022ASVspoof 2021: Towards Spoofed and Deepfake Speech Detection in the Wildcited by 118
- 2021Learning Debiased Representation via Disentangled Feature Augmentationcited by 116
- 2022Automatic speaker verification spoofing and deepfake detection using wav2vec 2.0 and data augmentationcited by 110
- 2017Towards Understanding Generalization of Deep Learning: Perspective of Loss Landscapescited by 84
- 2021RawBoost: A Raw Data Boosting and Augmentation Method applied to Automatic Speaker Verification Anti-Spoofingcited by 64
14 more
- 2020One-class Learning Towards Synthetic Voice Spoofing Detectioncited by 59
- 2021Speech is Silver, Silence is Golden: What do ASVspoof-trained Models Really Learn?cited by 36
- 2022SelecMix: Debiased Learning by Contradicting-pair Samplingcited by 29
- 2021Continual Learning for Fake Audio Detectioncited by 27
- 2023Can large-scale vocoded spoofed data improve speech spoofing countermeasure with a self-supervised front end?cited by 17
- 2021A Study On Data Augmentation In Voice Anti-Spoofingcited by 16
- 2023The Impact of Silence on Speech Anti-Spoofingcited by 13
- 2019Generalization of Spoofing Countermeasures: a Case Study with ASVspoof 2015 and BTAS 2016 Corporacited by 9
- 2023Multi-Dataset Co-Training with Sharpness-Aware Optimization for Audio Anti-spoofingcited by 9
- 2021Data Quality as Predictor of Voice Anti-Spoofing Generalizationcited by 6
- 2019Replay attack detection with complementary high-resolution information using end-to-end DNN for the ASVspoof 2019 Challengecited by 5
- 2020Dataset artefacts in anti-spoofing systems: a case study on the ASVspoof 2017 benchmarkcited by 5
- 2021Data Augmentation with Signal Companding for Detection of Logical Access Attackscited by 5
- 2022Investigating Active-learning-based Training Data Selection for Speech Spoofing Countermeasurecited by 5
Built on by
Citing papers per year: 2024: 1; 2025: 1.
2 papers.