Research map / Adversarial robustness and security / Backdoor and poisoning
Hidden in the Noise: Unveiling Backdoors in Audio LLMs Alignment Through Latent Acoustic Pattern Triggers
Liang Lin, Miao Yu, Kaiwen Luo, Yibo Zhang, Lilan Peng, Dexian Wang, Xuehai Tang, Yuanhe Zhang, Xikang Yang, Zhenhong Zhou, Kun Wang, Yang Liu
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 |
|---|---|---|---|
| use speed trigger for backdoor poisoning | ALLM backdoor · AudioSafe | ASR: 1.000 average ASR | Improved |
| use emotion trigger for backdoor poisoning | ALLM backdoor · AudioSafe | ASR: 1.000 average ASR | Improved |
| use volume as backdoor trigger | ALLM backdoor · AudioSafe | ASR: 0.049 average ASR | Hurt |
| use accent trigger instead of noise trigger | ALLM backdoor · AudioSafe | ASR: 0.511 vs 0.887 average ASR | Mixed |
| raise poisoning ratio for emotion trigger (ablation) | ALLM backdoor · AudioSafe | ASR | Improved |
| raise poisoning ratio for volume trigger (ablation) | ALLM backdoor · AudioSafe | ASR | No clear effect |
| raise poisoning ratio for accent trigger (ablation) | ALLM backdoor · AudioSafe | ASR | Improved |
| add Silero-VAD preprocessing defense | noise trigger · AudioSafe | ASR: 0.887 -> 0.013 | Improved |
| add Silero-VAD preprocessing defense | speed trigger · AudioSafe | ASR: 1.000 -> 0.833 | No clear effect |
| add Silero-VAD preprocessing defense | emotion trigger · AudioSafe | ASR: 1.000 -> 0.763 | No clear effect |
| add Fine-Mixing defense | emotion trigger · AudioSafe | ASR: 1.000 -> 0.016 | Improved |
| add Fine-Mixing defense | Qwen2.5-Omni · AudioSafe | ACC | Hurt |
Closest papers by meaning
- ALMGuard: Safety Shortcuts and Where to Find Them as Guardrails for Audio–Language Models (NeurIPS 2025 · similarity 0.89)
- Adaptive Compression and Targeted Perturbation: A Unified Framework for Generalized Audio Deepfake Detection (NeurIPS 2026 · similarity 0.86)
- Jailbreak-AudioBench: In-Depth Evaluation and Analysis of Jailbreak Threats for Large Audio Language Models (NeurIPS 2025 · similarity 0.86)
- Speed Master: Quick or Slow Play to Attack Speaker Recognition (AAAI 2025 · similarity 0.86)
- Rethinking Audio-Visual Adversarial Vulnerability from Temporal and Modality Perspectives (ICLR 2025 · similarity 0.86)
- AdvWave: Stealthy Adversarial Jailbreak Attack against Large Audio-Language Models (ICLR 2025 · similarity 0.86)
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.
