Audio Jailbreak Attacks: Exposing Vulnerabilities in SpeechGPT in a White-Box Framework
cs.CL
Submitted: 2025-05-24
Updated: 2025-05-24
Journal ref: 2025 55th Annual IEEE/IFIP International Conference on Dependable Systems and Networks Workshops (DSN-W), pp. 259-265 (2025)
DOI: 10.1109/DSN-W65791.2025.00074
Code: https://github.com/Magic-Ma-tech/AudioJailbreak-Attacks
License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Terminology
Sources
- GPT-4 Technical Report
- SpeechGPT: Empowering Large Language Models with Intrinsic Cross-Modal Conversational Abilities
- GPT-4o System Card
- Voice Jailbreak Attacks Against GPT-4o
- Jailbreaking Black Box Large Language Models in Twenty Queries
- AutoDAN: Generating Stealthy Jailbreak Prompts on Aligned Large Language Models
- Universal and Transferable Adversarial Attacks on Aligned Language Models
- JailbreakRadar: Comprehensive Assessment of Jailbreak Attacks Against LLMs
- Catastrophic Jailbreak of Open-source LLMs via Exploiting Generation
- FigStep: Jailbreaking Large Vision-Language Models via Typographic Visual Prompts
- Multi-step Jailbreaking Privacy Attacks on ChatGPT
- HarmBench: A Standardized Evaluation Framework for Automated Red Teaming and Robust Refusal
- A StrongREJECT for Empty Jailbreaks
- GPTFUZZER: Red Teaming Large Language Models with Auto-Generated Jailbreak Prompts
- Multilingual Jailbreak Challenges in Large Language Models
- Low-Resource Languages Jailbreak GPT-4
- SmoothLLM: Defending Large Language Models Against Jailbreaking Attacks
- Probing the Robustness of Vision-Language Pretrained Models: A Multimodal Adversarial Attack Approach
- Transferability of Adversarial Attacks in Video-based MLLMs: A Cross-modal Image-to-Video Approach
- Speech Resynthesis from Discrete Disentangled Self-Supervised Representations
Related papers
- ALARM: Audio-Language Alignment for Reasoning Models
- ThinkSum: Probabilistic reasoning over sets using large language models
- Coherence boosting: When your pretrained language model is not paying enough attention
- From Facts to Insights: A Persona-Driven Dual Memory Framework and Dataset for Role-Playing Agents
- Improving General Role-Playing Agents via Psychology-Grounded Reasoning and Role-Aware Policy Optimization
- The Art of Asking: Multilingual Prompt Optimization for Synthetic Data