PulseAugur
EN
LIVE 22:33:07

New attack bypasses audio LLM defenses via latent space manipulation

Researchers have developed a novel adversarial attack called CodecAttack that bypasses compression defenses in audio large language models (LLMs). Unlike previous methods that directly perturbed audio waveforms, CodecAttack optimizes perturbations within a neural audio codec's latent space. This technique proved highly effective, achieving an average 85.5% attack success rate on Opus compression and demonstrating transferability to other codecs like MP3 and AAC-LC. AI

IMPACT Demonstrates that lossy compression is not a reliable defense against adversarial audio, posing a practical threat to deployed audio LLM systems.

RANK_REASON Academic paper detailing a new method for attacking AI models. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.AI →

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →

COVERAGE [1]

  1. arXiv cs.AI TIER_1 English(EN) · Jaechul Roh, Jean-Philippe Monteuuis, Jonathan Petit, Amir Houmansdar ·

    Codec-Robust Attacks on Audio LLMs

    arXiv:2605.20519v1 Announce Type: cross Abstract: Prior attacks on Audio Large Language Models (Audio LLMs) demonstrated that carefully crafted waveform-domain perturbations can force targeted adversarial outputs. As a defense mechanism against these attacks, real-world codec com…