Imagen de Portada

Compiling Activation Steering into Weights via Null-Space Constraints for Stealthy Backdoors

Fuente: arXiv

Guardado en:

Detalles Bibliográficos
Autores principales:	Yin, Rui, Han, Tianxu, Xu, Naen, Li, Changjiang, He, Ping, Zhou, Chunyi, Wang, Jun, Fu, Zhihui, Du, Tianyu, Li, Jinbao, Ji, Shouling
Formato:	Preprint
Publicado:	2026
Materias:	Cryptography and Security Computation and Language
Acceso en línea:	Acceder al recurso
Etiquetas:	Agregar Etiqueta Sin Etiquetas, Sea el primero en etiquetar este registro!

Ejemplares similares

When Agents "Misremember" Collectively: Exploring the Mandela Effect in LLM-based Multi-Agent Systems
por: Xu, Naen, et al.
Publicado: (2026)

VideoEraser: Concept Erasure in Text-to-Video Diffusion Models
por: Xu, Naen, et al.
Publicado: (2025)

ACIArena: Toward Unified Evaluation for Agent Cascading Injection
por: An, Hengyu, et al.
Publicado: (2026)

FraudShield: Knowledge Graph Empowered Defense for LLMs against Fraud Attacks
por: Xu, Naen, et al.
Publicado: (2026)

Dullahan: Stealthy Backdoor Attack against Without-Label-Sharing Split Learning
por: Pu, Yuwen, et al.
Publicado: (2024)

Bridging the Copyright Gap: Do Large Vision-Language Models Recognize and Respect Copyrighted Content?
por: Xu, Naen, et al.
Publicado: (2025)

Automatic Red Teaming LLM-based Agents with Model Context Protocol Tools
por: He, Ping, et al.
Publicado: (2025)

"I See What You Did There": Can Large Vision-Language Models Understand Multimodal Puns?
por: Xu, Naen, et al.
Publicado: (2026)

Enhancing Adversarial Transferability with Adversarial Weight Tuning
por: Chen, Jiahao, et al.
Publicado: (2024)

Watch the Watcher! Backdoor Attacks on Security-Enhancing Diffusion Models
por: Li, Changjiang, et al.
Publicado: (2024)

CLIBE: Detecting Dynamic Backdoors in Transformer-based NLP Models
por: Zeng, Rui, et al.
Publicado: (2024)

UNIDOOR: A Universal Framework for Action-Level Backdoor Attacks in Deep Reinforcement Learning
por: Ma, Oubo, et al.
Publicado: (2025)

Mellivora Capensis: A Backdoor-Free Training Framework on the Poisoned Dataset without Auxiliary Data
por: Pu, Yuwen, et al.
Publicado: (2024)

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

CAMH: Advancing Model Hijacking Attack in Machine Learning
por: He, Xing, et al.
Publicado: (2024)

Rethinking the Vulnerabilities of Face Recognition Systems:From a Practical Perspective
por: Chen, Jiahao, et al.
Publicado: (2024)

Under-confidence Backdoors Are Resilient and Stealthy Backdoors
por: Peng, Minlong, et al.
Publicado: (2022)

Angel or Demon: Investigating the Plasticity Interventions' Impact on Backdoor Threats in Deep Reinforcement Learning
por: Ma, Oubo, et al.
Publicado: (2026)

IPIGuard: A Novel Tool Dependency Graph-Based Defense Against Indirect Prompt Injection in LLM Agents
por: An, Hengyu, et al.
Publicado: (2025)

Towards Backdoor Stealthiness in Model Parameter Space
por: Xu, Xiaoyun, et al.
Publicado: (2025)

Poison in the Well: Feature Embedding Disruption in Backdoor Attacks
por: Feng, Zhou, et al.
Publicado: (2025)

DOPA: Stealthy and Generalizable Backdoor Attacks from a Single Client under Challenging Federated Constraints
por: Qin, Xuezheng, et al.
Publicado: (2025)

Stealthy Targeted Backdoor Attacks against Image Captioning
por: Fan, Wenshu, et al.
Publicado: (2024)

Large Language Models are Good Attackers: Efficient and Stealthy Textual Backdoor Attacks
por: Li, Ziqiang, et al.
Publicado: (2024)

"No Matter What You Do": Purifying GNN Models via Backdoor Unlearning
por: Zhang, Jiale, et al.
Publicado: (2024)

On the Security Risks of ML-based Malware Detection Systems: A Survey
por: He, Ping, et al.
Publicado: (2025)

Profiling for Pennies: Unveiling the Privacy Iceberg of LLM Agents
por: Chen, Jiahao, et al.
Publicado: (2026)

State Backdoor: Towards Stealthy Real-world Poisoning Attack on Vision-Language-Action Model in State Space
por: Guo, Ji, et al.
Publicado: (2026)

CopyrightMeter: Revisiting Copyright Protection in Text-to-image Models
por: Xu, Naen, et al.
Publicado: (2024)

Towards Effective, Stealthy, and Persistent Backdoor Attacks Targeting Graph Foundation Models
por: Luo, Jiayi, et al.
Publicado: (2025)

Cuckoo Attack: Stealthy and Persistent Attacks Against AI-IDE
por: Liu, Xinpeng, et al.
Publicado: (2025)

LoRAShield: Data-Free Editing Alignment for Secure Personalized LoRA Sharing
por: Chen, Jiahao, et al.
Publicado: (2025)

TooBadRL: Trigger Optimization to Boost Effectiveness of Backdoor Attacks on Deep Reinforcement Learning
por: Zhang, Mingxuan, et al.
Publicado: (2025)

Stealthy Backdoor Attack via Confidence-driven Sampling
por: He, Pengfei, et al.
Publicado: (2023)

The Eminence in Shadow: Exploiting Feature Boundary Ambiguity for Robust Backdoor Attacks
por: Feng, Zhou, et al.
Publicado: (2025)

Exposing Vulnerabilities in RL: A Novel Stealthy Backdoor Attack through Reward Poisoning
por: Zhang, Bokang, et al.
Publicado: (2025)

MirageBackdoor: A Stealthy Attack that Induces Think-Well-Answer-Wrong Reasoning
por: Zeng, Yizhe, et al.
Publicado: (2026)

BadActs: A Universal Backdoor Defense in the Activation Space
por: Yi, Biao, et al.
Publicado: (2024)

Unveiling the Security Risks of Federated Learning in the Wild: From Research to Practice
por: Chen, Jiahao, et al.
Publicado: (2026)

Towards Stealthy and Effective Backdoor Attacks on Lane Detection: A Naturalistic Data Poisoning Approach
por: Liao, Yifan, et al.
Publicado: (2025)