@LeeLeepenkman: Text-Anchored Semantic Perturbations for Transferable Jailbreak Attacks on Multimodal Large Language Models https://pap…
Summary
The paper introduces TA-SPA, a black-box jailbreak attack framework for multimodal large language models that uses text-anchored semantic perturbations to achieve effective and transferable attacks against safety alignments.
View Cached Full Text
Cached at: 08/26/26, 07:24 AM
Text-Anchored Semantic Perturbations for Transferable Jailbreak Attacks on Multimodal Large Language Models
https://t.co/NljjVIHrnS https://t.co/NGrf1icnF1
Text-Anchored Semantic Perturbations for Transferable Jailbreak Attacks on Multimodal Large Language Models
Source: https://papers.app.nz/view/paper?id=9107387 Multimodal Large Language Models (MLLMs) have achieved remarkable progress in vision-language interaction, yet their safety alignment remains vulnerable to jailbreak attacks. A key challenge is that safety behavior learned in the textual space does not reliably transfer to fused cross-modal representations, leaving multimodal inputs exploitable through latent semantic cues. We propose Text-Anchored Semantic Perturbation Attack (TA-SPA), a black-box jailbreak framework that optimizes transferable perturbations in a text-anchored semantic space. TA-SPA integrates Text-Anchored Semantic Factorization (TASF), which encourages the separation of cross-modal semantic factors from modality-specific residuals, with Semantic-Preserving Augmentation (SPA), which diversifies harmful target anchors while preserving semantic consistency. Experiments show strong attack effectiveness and transfer to commercial MLLMs, with competitive performance under representative defenses. Additional controls and probing support the intended factorization without implying perfect disentanglement, motivating representation-level safety alignment beyond input-level filtering.
Similar Articles
Same Model, Different Weakness: How Language and Modality Reshape the Jailbreak Attack Surface in Frontier MLLMs
This paper presents the first systematic cross-lingual, multimodal red-teaming study comparing jailbreak vulnerability in US English and Mexican Spanish across four frontier MLLMs, revealing that language does not scale vulnerability uniformly and that safety rankings are not preserved across languages.
One Jailbreak, Many Tongues: Learning Language-Insensitive Intention Representations for Multilingual Jailbreak Detection
This paper proposes MLJailDe, a multilingual jailbreak detection framework that uses back-translation data augmentation and relative-distance constraints to improve cross-lingual generalization and robustness, achieving 98.5% F1 score across 11 languages.
Evolving Safety Landscape of Multi-modal Large Language Models: A Survey of Emerging Threats and Safeguards
A survey paper systematically analyzing the evolving safety landscape of multi-modal large language models, covering emerging threats such as adversarial attacks, data poisoning, jailbreaks, and hallucinations, and reviewing updated safety strategies.
Conceal, Reconstruct, Jailbreak: Exploiting the Reconstruction-Concealment Tradeoff in MLLMs
This paper analyzes the reconstruction-concealment tradeoff in intent-obfuscation jailbreak attacks on Multimodal Large Language Models (MLLMs). It proposes concealment-aware variant construction and keyword-related distractor images to exploit model vulnerabilities more effectively.
An Empirical Study of Multi-Generation Sampling for Jailbreak Detection in Large Language Models
Empirical study shows multi-generation sampling significantly improves jailbreak detection in LLMs, revealing hidden harmful outputs that single-generation audits miss.