Copertina

Spot Risks Before Speaking! Unraveling Safety Attention Heads in Large Vision-Language Models

Fuente: arXiv

Salvato in:

Dettagli Bibliografici
Autori principali:	Zheng, Ziwei, Zhao, Junyao, Yang, Le, He, Lijun, Li, Fan
Natura:	Preprint
Pubblicazione:	2025
Soggetti:	Machine Learning Artificial Intelligence Cryptography and Security Computer Vision and Pattern Recognition
Accesso online:	Acceder al recurso
Tags:	Aggiungi Tag Nessun Tag, puoi essere il primo ad aggiungerne!!

Documenti analoghi

Test-Time Attention Purification for Backdoored Large Vision Language Models
di: Zhang, Zhifang, et al.
Pubblicazione: (2026)

Rethinking Bottlenecks in Safety Fine-Tuning of Vision Language Models
di: Ding, Yi, et al.
Pubblicazione: (2025)

DAVSP: Safety Alignment for Large Vision-Language Models via Deep Aligned Visual Safety Prompt
di: Zhang, Yitong, et al.
Pubblicazione: (2025)

A Survey of Safety on Large Vision-Language Models: Attacks, Defenses and Evaluations
di: Ye, Mang, et al.
Pubblicazione: (2025)

R-TPT: Improving Adversarial Robustness of Vision-Language Models through Test-Time Prompt Tuning
di: Sheng, Lijun, et al.
Pubblicazione: (2025)

AGMark: Attention-Guided Dynamic Watermarking for Large Vision-Language Models
di: Li, Yue, et al.
Pubblicazione: (2026)

Revisiting Data Auditing in Large Vision-Language Models
di: Zhu, Hongyu, et al.
Pubblicazione: (2025)

Inducing High Energy-Latency of Large Vision-Language Models with Verbose Images
di: Gao, Kuofeng, et al.
Pubblicazione: (2024)

A Cross-Modal Prompt Injection Attack against Large Vision-Language Models with Image-Only Perturbation
di: Yang, Hao, et al.
Pubblicazione: (2026)

Image-Based Geolocation Using Large Vision-Language Models
di: Liu, Yi, et al.
Pubblicazione: (2024)

Backdoor Attack on Vision Language Models with Stealthy Semantic Manipulation
di: Zhong, Zhiyuan, et al.
Pubblicazione: (2025)

Hidden Tail: Adversarial Image Causing Stealthy Resource Consumption in Vision-Language Models
di: Zhang, Rui, et al.
Pubblicazione: (2025)

Unbridled Icarus: A Survey of the Potential Perils of Image Inputs in Multimodal Large Language Model Security
di: Fan, Yihe, et al.
Pubblicazione: (2024)

Image Corruption-Inspired Membership Inference Attacks against Large Vision-Language Models
di: Wu, Zongyu, et al.
Pubblicazione: (2025)

Evolving Contextual Safety in Multi-Modal Large Language Models via Inference-Time Self-Reflective Memory
di: Zhang, Ce, et al.
Pubblicazione: (2026)

Making Every Step Effective: Jailbreaking Large Vision-Language Models Through Hierarchical KV Equalization
di: Hao, Shuyang, et al.
Pubblicazione: (2025)

MultiPriv: Benchmarking Individual-Level Privacy Reasoning in Vision-Language Models
di: Sun, Xiongtao, et al.
Pubblicazione: (2025)

Text is All You Need for Vision-Language Model Jailbreaking
di: Chen, Yihang, et al.
Pubblicazione: (2026)

VLM-Guard: Safeguarding Vision-Language Models via Fulfilling Safety Alignment Gap
di: Liu, Qin, et al.
Pubblicazione: (2025)

Light as Deception: GPT-driven Natural Relighting Against Vision-Language Pre-training Models
di: Yang, Ying, et al.
Pubblicazione: (2025)

Mirage in the Eyes: Hallucination Attack on Multi-modal Large Language Models with Only Attention Sink
di: Wang, Yining, et al.
Pubblicazione: (2025)

Similarity Distribution based Membership Inference Attack on Person Re-identification
di: Gao, Junyao, et al.
Pubblicazione: (2022)

Membership Inference Attack Against Masked Image Modeling
di: Li, Zheng, et al.
Pubblicazione: (2024)

Depth Charge: Jailbreak Large Language Models from Deep Safety Attention Heads
di: Wu, Jinman, et al.
Pubblicazione: (2026)

Jailbreak Vision Language Models via Bi-Modal Adversarial Prompt
di: Ying, Zonghao, et al.
Pubblicazione: (2024)

FIDAVL: Fake Image Detection and Attribution using Vision-Language Model
di: Keita, Mamadou, et al.
Pubblicazione: (2024)

ViTGuard: Attention-aware Detection against Adversarial Examples for Vision Transformer
di: Sun, Shihua, et al.
Pubblicazione: (2024)

AttackVLA: Benchmarking Adversarial and Backdoor Attacks on Vision-Language-Action Models
di: Li, Jiayu, et al.
Pubblicazione: (2025)

Harnessing the Power of Large Vision Language Models for Synthetic Image Detection
di: Keita, Mamadou, et al.
Pubblicazione: (2024)

Learning to Detect Unseen Jailbreak Attacks in Large Vision-Language Models
di: Liang, Shuang, et al.
Pubblicazione: (2025)

Blind Spot Navigation: Evolutionary Discovery of Sensitive Semantic Concepts for LVLMs
di: Pan, Zihao, et al.
Pubblicazione: (2025)

ViT-EnsembleAttack: Augmenting Ensemble Models for Stronger Adversarial Transferability in Vision Transformers
di: Cao, Hanwen, et al.
Pubblicazione: (2025)

On the Adversarial Robustness of Large Vision-Language Models under Visual Token Compression
di: Zhang, Xinwei, et al.
Pubblicazione: (2026)

VLATTACK: Multimodal Adversarial Attacks on Vision-Language Tasks via Pre-trained Models
di: Yin, Ziyi, et al.
Pubblicazione: (2023)

PromptSmooth: Certifying Robustness of Medical Vision-Language Models via Prompt Learning
di: Hussein, Noor, et al.
Pubblicazione: (2024)

Do Not Leave a Gap: Hallucination-Free Object Concealment in Vision-Language Models
di: Guesmi, Amira, et al.
Pubblicazione: (2026)

Self-adaptive Dataset Construction for Real-World Multimodal Safety Scenarios
di: Qu, Jingen, et al.
Pubblicazione: (2025)

When Understanding Becomes a Risk: Authenticity and Safety Risks in the Emerging Image Generation Paradigm
di: Leng, Ye, et al.
Pubblicazione: (2026)

One Perturbation is Enough: On Generating Universal Adversarial Perturbations against Vision-Language Pre-training Models
di: Fang, Hao, et al.
Pubblicazione: (2024)

$\mathbf{S^2LM}$: Towards Semantic Steganography via Large Language Models
di: Wu, Huanqi, et al.
Pubblicazione: (2025)