Research map / Adversarial robustness and security / Attacks on text and graphs
Time Shuffle: A Transferability-Booster for Multiple Audio Adversarial Tasks
JiaCheng Deng, Dengpan Ye, Yuhong Liu, Zhaolin Wei, Ziyi Liu, Haoran Duan
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 TS-p (phoneme-level time shuffle) to I-FGSM (ablation) | I-FGSM · ASR conformer | WER: 0.124 -> 0.24 | Improved |
| add TS-p to DI-FGSM (ablation) | DI-FGSM · ASR wav2vec (white-box) | WER: 0.552 -> 1.36 | Improved |
| add TS-p to RMA (ablation) | RMA · ASR conformer | WER: 0.108 -> 0.28 | Improved |
| add TS-p to STA-MDCT (ablation) | STA-MDCT · ASR wav2vec | WER: 0.842 -> 0.03 | Hurt |
| add TS-p to I-FGSM (ablation) | I-FGSM · KWS wav kws | TASR: 0.43 -> 0.56 | Improved |
| add TS-p to DI-FGSM (ablation) | DI-FGSM · ASV ecapa-tdnn | cosine similarity: 0.736 -> 0.63 | Improved |
| add TS-p to attacks, average across tasks (black-box) (ablation) | TS-p series · ASR/ASV/KWS | average black-box transfer: 0.173 -> 0.213 ASR; 0.660 -> 0.61 ASV; 0.47 -> 0.50 KWS | Improved |
| phoneme-level instead of segment-level shuffle (ablation) | Time Shuffle · ASV | cosine similarity: 0.643 -> 0.599 | Improved |
| increase shuffle gradient weight w to 1 (ablation) | TS-p · ASR/ASV/KWS | transfer attack metric: -0.007, -0.004, -0.115 black-box | Hurt |
| increase number of shuffles M (ablation) | TS-p · ASR, ASV vs KWS | transfer attack metric | Mixed |
| TS-p3 vs best baseline on white-box wav2vec | TS-p3 · ASR wav2vec | WER: 1.438 -> 1.697 | Improved |
| TS-p3 vs ACG on crdnn | TS-p3 · ASR crdnn | WER: 0.302 -> 0.263 | Hurt |
Closest papers by meaning
- Rethinking Audio-Visual Adversarial Vulnerability from Temporal and Modality Perspectives (ICLR 2025 · similarity 0.89)
- Breaking Adversarial Transferability in Fine-Tuned Speech Recognition (NeurIPS 2026 · similarity 0.87)
- AdvWave: Stealthy Adversarial Jailbreak Attack against Large Audio-Language Models (ICLR 2025 · similarity 0.86)
- Speed Master: Quick or Slow Play to Attack Speaker Recognition (AAAI 2025 · similarity 0.85)
- ALMGuard: Safety Shortcuts and Where to Find Them as Guardrails for Audio–Language Models (NeurIPS 2025 · similarity 0.84)
- Adaptive Compression and Targeted Perturbation: A Unified Framework for Generalized Audio Deepfake Detection (NeurIPS 2026 · 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.
