Breaking Audio Large Language Models by Attacking Only the Encoder: A Universal Targeted Latent-Space Audio Attack
Fuente:
arXiv
Salvato in:
| Autori principali: | , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2025
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866914225050353664 |
|---|---|
| author | Ziv, Roee Lapid, Raz Sipper, Moshe |
| author_facet | Ziv, Roee Lapid, Raz Sipper, Moshe |
| contents | Audio-language models combine audio encoders with large language models to enable multimodal reasoning, but they also introduce new security vulnerabilities. We propose a universal targeted latent space attack, an encoder-level adversarial attack that manipulates audio latent representations to induce attacker-specified outputs in downstream language generation. Unlike prior waveform-level or input-specific attacks, our approach learns a universal perturbation that generalizes across inputs and speakers and does not require access to the language model. Experiments on Qwen2-Audio-7B-Instruct demonstrate consistently high attack success rates with minimal perceptual distortion, revealing a critical and previously underexplored attack surface at the encoder level of multimodal systems. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2512_23881 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Breaking Audio Large Language Models by Attacking Only the Encoder: A Universal Targeted Latent-Space Audio Attack Ziv, Roee Lapid, Raz Sipper, Moshe Sound Artificial Intelligence Cryptography and Security Audio-language models combine audio encoders with large language models to enable multimodal reasoning, but they also introduce new security vulnerabilities. We propose a universal targeted latent space attack, an encoder-level adversarial attack that manipulates audio latent representations to induce attacker-specified outputs in downstream language generation. Unlike prior waveform-level or input-specific attacks, our approach learns a universal perturbation that generalizes across inputs and speakers and does not require access to the language model. Experiments on Qwen2-Audio-7B-Instruct demonstrate consistently high attack success rates with minimal perceptual distortion, revealing a critical and previously underexplored attack surface at the encoder level of multimodal systems. |
| title | Breaking Audio Large Language Models by Attacking Only the Encoder: A Universal Targeted Latent-Space Audio Attack |
| topic | Sound Artificial Intelligence Cryptography and Security |
| url | https://arxiv.org/abs/2512.23881 |