Image de couverture de livre

Servant, Stalker, Predator: How An Honest, Helpful, And Harmless (3H) Agent Unlocks Adversarial Skills

Fuente: arXiv

Enregistré dans:

Détails bibliographiques
Auteur principal:	Noever, David
Format:	Preprint
Publié:	2025
Sujets:	Cryptography and Security Artificial Intelligence
Accès en ligne:	Acceder al recurso
Tags:	Ajouter un tag Pas de tags, Soyez le premier à ajouter un tag!

Documents similaires

Language Models And A Second Opinion Use Case: The Pocket Professional
par: Noever, David
Publié: (2024)

Hallucinating AI Hijacking Attack: Large Language Models and Malicious Code Recommenders
par: Noever, David, et autres
Publié: (2024)

Towards Safe and Honest AI Agents with Neural Self-Other Overlap
par: Carauleanu, Marc, et autres
Publié: (2024)

HarmfulSkillBench: How Do Harmful Skills Weaponize Your Agents?
par: Jiang, Yukun, et autres
Publié: (2026)

Attention Masks Help Adversarial Attacks to Bypass Safety Detectors
par: Shi, Yunfan
Publié: (2024)

Understanding LLM Behavior When Encountering User-Supplied Harmful Content in Harmless Tasks
par: Chu, Junjie, et autres
Publié: (2026)

HonestCyberEval: An AI Cyber Risk Benchmark for Automated Software Exploitation
par: Ristea, Dan, et autres
Publié: (2024)

SkillTester: Benchmarking Utility and Security of Agent Skills
par: Wang, Leye, et autres
Publié: (2026)

BadSkill: Backdoor Attacks on Agent Skills via Model-in-Skill Poisoning
par: Tie, Guiyao, et autres
Publié: (2026)

SkillJect: Effectively Automating Skill-Based Prompt Injection for Skill-Enabled Agents
par: Jia, Xiaojun, et autres
Publié: (2026)

SkillTrojan: Backdoor Attacks on Skill-Based Agent Systems
par: Feng, Yunhao, et autres
Publié: (2026)

Context Matters: Repository-Aware Security Analysis of the Agent Skill Ecosystem
par: Holzbauer, Florian, et autres
Publié: (2026)

SkillSieve: A Hierarchical Triage Framework for Detecting Malicious AI Agent Skills
par: Hou, Yinghan, et autres
Publié: (2026)

AgentTrap: Measuring Runtime Trust Failures in Third-Party Agent Skills
par: Zhuang, Haomin, et autres
Publié: (2026)

Detecting Adversarial Fine-tuning with Auditing Agents
par: Egler, Sarah, et autres
Publié: (2025)

Technical Report: Exploring the Emerging Threats of the Agent Skill Ecosystem
par: Beurer-Kellner, Luca, et autres
Publié: (2026)

Structured Security Auditing and Robustness Enhancement for Untrusted Agent Skills
par: Lv, Lijia, et autres
Publié: (2026)

No Attack Required: Semantic Fuzzing for Specification Violations in Agent Skills
par: Li, Ying, et autres
Publié: (2026)

Living Off the LLM: How LLMs Will Change Adversary Tactics
par: Oesch, Sean, et autres
Publié: (2025)

Towards Secure Agent Skills: Architecture, Threat Taxonomy, and Security Analysis
par: Li, Zhiyuan, et autres
Publié: (2026)

Proteus: A Self-Evolving Red Team for Agent Skill Ecosystems
par: Zhou, Zhaojiacheng
Publié: (2026)

RouteGuard: Internal-Signal Detection of Skill Poisoning in LLM Agents
par: Xiao, Wenjie, et autres
Publié: (2026)

Latent Fusion Jailbreak: Blending Harmful and Harmless Representations to Elicit Unsafe LLM Outputs
par: Xing, Wenpeng, et autres
Publié: (2025)

Adversarial Agents: Black-Box Evasion Attacks with Reinforcement Learning
par: Domico, Kyle, et autres
Publié: (2025)

SkCC: Portable and Secure Skill Compilation for Cross-Framework LLM Agents
par: Ouyang, Yipeng, et autres
Publié: (2026)

Credential Leakage in LLM Agent Skills: A Large-Scale Empirical Study
par: Chen, Zhihao, et autres
Publié: (2026)

WARD: Adversarially Robust Defense of Web Agents Against Prompt Injections
par: Cao, Tri, et autres
Publié: (2026)

How stealthy is stealthy? Studying the Efficacy of Black-Box Adversarial Attacks in the Real World
par: Panebianco, Francesco, et autres
Publié: (2025)

ClawKeeper: Comprehensive Safety Protection for OpenClaw Agents Through Skills, Plugins, and Watchers
par: Liu, Songyang, et autres
Publié: (2026)

Under the Hood of SKILL.md: Semantic Supply-chain Attacks on AI Agent Skill Registry
par: Saha, Shoumik, et autres
Publié: (2026)

DECEPTICON: How Dark Patterns Manipulate Web Agents
par: Cuvin, Phil, et autres
Publié: (2025)

Explanation as a Watermark: Towards Harmless and Multi-bit Model Ownership Verification via Watermarking Feature Attribution
par: Shao, Shuo, et autres
Publié: (2024)

Doppelganger Method: Breaking Role Consistency in LLM Agent via Prompt-based Transferable Adversarial Attack
par: Kang, Daewon, et autres
Publié: (2025)

When Safe Skills Collide: Measuring Compositional Risk in Agent Skill Ecosystems
par: Wang, Su, et autres
Publié: (2026)

The Blind Spot of Agent Safety: How Benign User Instructions Expose Critical Vulnerabilities in Computer-Use Agents
par: Ding, Xuwei, et autres
Publié: (2026)

Next-Generation Phishing: How LLM Agents Empower Cyber Attackers
par: Afane, Khalifa, et autres
Publié: (2024)

Behavioral Integrity Verification for AI Agent Skills
par: Wu, Yuhao, et autres
Publié: (2026)

AccLock: Unlocking Identity with Heartbeat Using In-Ear Accelerometers
par: Wang, Lei, et autres
Publié: (2026)

From Allies to Adversaries: Manipulating LLM Tool-Calling through Adversarial Injection
par: Wang, Haowei, et autres
Publié: (2024)

E-PhishGen: Unlocking Novel Research in Phishing Email Detection
par: Pajola, Luca, et autres
Publié: (2025)