Imagen de Portada

SafeSteer: Adaptive Subspace Steering for Efficient Jailbreak Defense in Vision-Language Models

Fuente: arXiv

Guardado en:

Detalles Bibliográficos
Autores principales:	Zeng, Xiyu, Liang, Siyuan, Lu, Liming, Zhu, Haotian, Liu, Enguang, Dang, Jisheng, Zhou, Yongbin, Pang, Shuchao
Formato:	Preprint
Publicado:	2025
Materias:	Cryptography and Security
Acceso en línea:	Acceder al recurso
Etiquetas:	Agregar Etiqueta Sin Etiquetas, Sea el primero en etiquetar este registro!

Ejemplares similares

Adaptive Probe-based Steering for Robust LLM Jailbreaking
por: Chen, Junxi, et al.
Publicado: (2026)

AgentRAE: Remote Action Execution through Notification-based Visual Backdoors against Screenshots-based Mobile GUI Agents
por: Luo, Yutao, et al.
Publicado: (2026)

DREAM: Dynamic Red-teaming across Environments for AI Models
por: Lu, Liming, et al.
Publicado: (2025)

AdaSteer: Your Aligned LLM is Inherently an Adaptive Jailbreak Defender
por: Zhao, Weixiang, et al.
Publicado: (2025)

Steering Externalities: Benign Activation Steering Unintentionally Increases Jailbreak Risk for Large Language Models
por: Xiong, Chen, et al.
Publicado: (2026)

Jailbreaking the Matrix: Nullspace Steering for Controlled Model Subversion
por: Pramanik, Vishal, et al.
Publicado: (2026)

PII Jailbreaking in LLMs via Activation Steering Reveals Personal Information Leakage
por: Nakka, Krishna Kanth, et al.
Publicado: (2025)

E$^2$AT: Multimodal Jailbreak Defense via Dynamic Joint Optimization for Multimodal Large Language Models
por: Lu, Liming, et al.
Publicado: (2025)

ShallowJail: Steering Jailbreaks against Large Language Models
por: Liu, Shang, et al.
Publicado: (2026)

AutoJailbreak: Exploring Jailbreak Attacks and Defenses through a Dependency Lens
por: Lu, Lin, et al.
Publicado: (2024)

SteerDiff: Steering towards Safe Text-to-Image Diffusion Models
por: Zhang, Hongxiang, et al.
Publicado: (2024)

TRACE: Task-Aware Adaptive Self-Evolving Agentic Jailbreaking
por: Zeng, Churui, et al.
Publicado: (2026)

T2VShield: Model-Agnostic Jailbreak Defense for Text-to-Video Models
por: Liang, Siyuan, et al.
Publicado: (2025)

Reconstruction of Differentially Private Text Sanitization via Large Language Models
por: Pang, Shuchao, et al.
Publicado: (2024)

CEE: An Inference-Time Jailbreak Defense for Embodied Intelligence via Subspace Concept Rotation
por: Yang, Jirui, et al.
Publicado: (2025)

SARSteer: Safeguarding Large Audio-Language Models via Safe-Ablated Refusal Steering
por: Lin, Weilin, et al.
Publicado: (2025)

Steering Dialogue Dynamics for Robustness against Multi-turn Jailbreaking Attacks
por: Hu, Hanjiang, et al.
Publicado: (2025)

FlowSteer: Prompt-Only Workflow Steering Exposes Planning-Time Vulnerabilities in Multi-Agent LLM Systems
por: Li, Fanxiao, et al.
Publicado: (2026)

From Storage to Steering: Memory Control Flow Attacks on LLM Agents
por: Xu, Zhenlin, et al.
Publicado: (2026)

AlphaSteer: Learning Refusal Steering with Principled Null-Space Constraint
por: Sheng, Leheng, et al.
Publicado: (2025)

Tit-for-Tat: Safeguarding Large Vision-Language Models Against Jailbreak Attacks via Adversarial Defense
por: Hao, Shuyang, et al.
Publicado: (2025)

Jailbreaking LLMs & VLMs: Mechanisms, Evaluation, and Unified Defense
por: Chen, Zejian, et al.
Publicado: (2026)

Enhanced MLLM Black-Box Jailbreaking Attacks and Defenses
por: Zhong, Xingwei, et al.
Publicado: (2025)

Defensive Prompt Patch: A Robust and Interpretable Defense of LLMs against Jailbreak Attacks
por: Xiong, Chen, et al.
Publicado: (2024)

SafeMobile: Chain-level Jailbreak Detection and Automated Evaluation for Multimodal Mobile Agents
por: Liang, Siyuan, et al.
Publicado: (2025)

ARGUS: Defending Against Multimodal Indirect Prompt Injection via Steering Instruction-Following Behavior
por: Lu, Weikai, et al.
Publicado: (2025)

CENSOR: Defense Against Gradient Inversion via Orthogonal Subspace Bayesian Sampling
por: Zhang, Kaiyuan, et al.
Publicado: (2025)

Revisiting JBShield: Breaking and Rebuilding Representation-Level Jailbreak Defenses
por: Derya, Kemal, et al.
Publicado: (2026)

TrapSuffix: Proactive Defense Against Adversarial Suffixes in Jailbreaking
por: Du, Mengyao, et al.
Publicado: (2026)

Retrieval-Augmented Defense: Adaptive and Controllable Jailbreak Prevention for Large Language Models
por: Yang, Guangyu, et al.
Publicado: (2025)

Prefix Guidance: A Steering Wheel for Large Language Models to Defend Against Jailbreak Attacks
por: Zhao, Jiawei, et al.
Publicado: (2024)

AutoDefense: Multi-Agent LLM Defense against Jailbreak Attacks
por: Zeng, Yifan, et al.
Publicado: (2024)

Constitutional Classifiers++: Efficient Production-Grade Defenses against Universal Jailbreaks
por: Cunningham, Hoagy, et al.
Publicado: (2026)

Rogue Cell: Adversarial Attack and Defense in Untrusted O-RAN Setup Exploiting the Traffic Steering xApp
por: Aizikovich, Eran, et al.
Publicado: (2025)

How Jailbreak Defenses Work and Ensemble? A Mechanistic Investigation
por: Long, Zhuohang, et al.
Publicado: (2025)

JailDAM: Jailbreak Detection with Adaptive Memory for Vision-Language Model
por: Nian, Yi, et al.
Publicado: (2025)

OmniSafeBench-MM: A Unified Benchmark and Toolbox for Multimodal Jailbreak Attack-Defense Evaluation
por: Jia, Xiaojun, et al.
Publicado: (2025)

Jailbreaking Generative AI: Multivector Phishing Threats and Transformer based Defenses
por: Mishra, Rina, et al.
Publicado: (2025)

Analysing the Safety Pitfalls of Steering Vectors
por: Li, Yuxiao, et al.
Publicado: (2026)

MASCing: Configurable Mixture-of-Experts Behavior via Activation Steering Masks
por: Lintelo, Jona te, et al.
Publicado: (2026)