Acoustic Interference: A New Paradigm Weaponizing Acoustic Latent Semantic for Universal Jailbreak against Large Audio Language Models

Abstract

The integration of audio modality into Large Audio Language Models (LALMs) significantly expands their attack surface. Existing jailbreak paradigms predominantly treat audio as a carrier for malicious payloads, relying on semantic optimization, acoustic parameter control, or additive perturbation to embed harmful content into the audio signal. In this work, we challenge this necessity and propose a new paradigm in which the role of audio shifts from content injection to safety alignment interfer

Categories

Framework mappings

Suggested from the entry's categories.

@misc{wang2026acoustic,
  title = {{Acoustic Interference: A New Paradigm Weaponizing Acoustic Latent Semantic for Universal Jailbreak against Large Audio Language Models}},
  author = {Yanyun Wang and Yu Huang and Zi Liang and Xixin Wu and Li Liu},
  year = {2026},
  month = may,
  eprint = {2605.18168},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2605.18168}
}