Research map / Adversarial robustness and security / Attacks on text and graphs
Learning Efficient Representations for Fake Speech Detection
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add residual connections to EfficientCNN (ablation) | EfficientCNN · ASVSpoof2019 LARGE | macro-F1: 94.14 -> 97.22 | Helped |
| add residual connections to EfficientCNN (ablation) | EfficientCNN · ASVSpoof2019 SMALL | macro-F1: 88.44 -> 86.91 | Hurt |
| add residual connections to EfficientCNN (ablation) | EfficientCNN · ASVSpoof2019 MEDIUM | macro-F1: 85.36 -> 86.05 | No clear effect |
| add multi-task source-prediction head (231 params) (ablation) | RES-EfficientCNN · ASVSpoof2019 SMALL/MEDIUM/LARGE | macro-F1: 86.91 -> 92.91; 86.05 -> 95.91; 97.22 -> 97.61 | Helped |
| add multi-task source-prediction head (231 params) (ablation) | EfficientCNN · ASVSpoof2019 MEDIUM | macro-F1: 85.36 -> 93.41 | Helped |
| add multi-task source-prediction head (231 params) (ablation) | EfficientCNN · ASVSpoof2019 SMALL/LARGE | macro-F1: 88.44 -> 88.44 (SMALL); 94.14 -> 86.00 (LARGE) | Mixed |
| scale model from SMALL to LARGE (ablation) | Multi-RES-EfficientCNN · ASVSpoof2019 | macro-F1: 92.91 -> 97.61 | Helped |
| fine-tune pretrained detector on 6.25% of new attack data (transfer learning) vs zero-shot | RES-EfficientCNN LARGE · RTVCSpoof | macro-F1: 91.35 -> 99.12 | Helped |
| multi-task training in low-data regimes (ablation) | RES-EfficientCNN · ASVSpoof2019 reduced training data | macro-F1 | Helped |
Closest papers by meaning
- Improving Generalization for AI-Synthesized Voice Detection (AAAI 2025 · similarity 0.88)
- I Can Hear You: Selective Robust Training for Deepfake Audio Detection (ICLR 2025 · similarity 0.87)
- Phoneme-Level Feature Discrepancies: A Key to Detecting Sophisticated Speech Deepfakes (AAAI 2025 · similarity 0.86)
- Breaking Adversarial Transferability in Fine-Tuned Speech Recognition (NeurIPS 2026 · similarity 0.85)
- AudioMarkBench: Benchmarking Robustness of Audio Watermarking (NeurIPS 2024 · similarity 0.85)
- SLIM: Style-Linguistics Mismatch Model for Generalized Audio Deepfake Detection (NeurIPS 2024 · similarity 0.84)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.