Image de couverture de livre

Why Are Web AI Agents More Vulnerable Than Standalone LLMs? A Security Analysis

Fuente: arXiv

Enregistré dans:

Détails bibliographiques
Auteurs principaux:	Chiang, Jeffrey Yang Fan, Lee, Seungjae, Huang, Jia-Bin, Huang, Furong, Chen, Yizheng
Format:	Preprint
Publié:	2025
Sujets:	Machine Learning Computation and Language
Accès en ligne:	Acceder al recurso
Tags:	Ajouter un tag Pas de tags, Soyez le premier à ajouter un tag!

Documents similaires

Your Agent is More Brittle Than You Think: Uncovering Indirect Injection Vulnerabilities in Agentic LLMs
par: Zhu, Wenhui, et autres
Publié: (2026)

Exploring ChatGPT for Next-generation Information Retrieval: Opportunities and Challenges
par: Huang, Yizheng, et autres
Publié: (2024)

Imagine, Verify, Execute: Memory-guided Agentic Exploration with Vision-Language Models
par: Lee, Seungjae, et autres
Publié: (2025)

When Models Know More Than They Say: Probing Analogical Reasoning in LLMs
par: McGovern, Hope, et autres
Publié: (2026)

More Than One Teacher: Adaptive Multi-Guidance Policy Optimization for Diverse Exploration
par: Yuan, Xiaoyang, et autres
Publié: (2025)

More Vulnerable than You Think: On the Stability of Tool-Integrated LLM Agents
par: Xiong, Weimin, et autres
Publié: (2025)

GATES: Self-Distillation under Privileged Context with Consensus Gating
par: Stein, Alex, et autres
Publié: (2026)

More Than a Score: Probing the Impact of Prompt Specificity on LLM Code Generation
par: Zi, Yangtian, et autres
Publié: (2025)

Mixture Compressor for Mixture-of-Experts LLMs Gains More
par: Huang, Wei, et autres
Publié: (2024)

Rethinking Layer Redundancy: Calibration Matters More Than Search in LLM Depth Pruning
par: Kim, Minkyu, et autres
Publié: (2026)

Why Are Linear RNNs More Parallelizable?
par: Merrill, William, et autres
Publié: (2026)

Why Does Self-Distillation (Sometimes) Degrade the Reasoning Capability of LLMs?
par: Kim, Jeonghye, et autres
Publié: (2026)

Towards Self-Improvement of LLMs via MCTS: Leveraging Stepwise Knowledge with Curriculum Preference Learning
par: Wang, Xiyao, et autres
Publié: (2024)

WEBSERV: A Full-Stack and RL-Ready Web Environment for Training Web Agents at Scale
par: Lu, Yuxuan, et autres
Publié: (2025)

MoreHopQA: More Than Multi-hop Reasoning
par: Schnitzler, Julian, et autres
Publié: (2024)

DynaFLIP: Rethinking Robotics Perception via Tri-Modal-Dynamics Guided Representation
par: Lee, Jusuk, et autres
Publié: (2026)

On the Security Vulnerabilities of Text-to-SQL Models
par: Peng, Xutan, et autres
Publié: (2022)

"More Than Words": Linking Music Preferences and Moral Values Through Lyrics
par: Preniqi, Vjosa, et autres
Publié: (2022)

More Than Bits: Multi-Envelope Double Binary Factorization for Extreme Quantization
par: Ichikawa, Yuma, et autres
Publié: (2025)

Enhancing Antibiotic Stewardship using a Natural Language Approach for Better Feature Representation
par: Lee, Simon A., et autres
Publié: (2024)

Clinical ModernBERT: An efficient and long context encoder for biomedical text
par: Lee, Simon A., et autres
Publié: (2025)

How Vulnerable Are Edge LLMs?
par: Ding, Ao, et autres
Publié: (2026)

Harder Tasks Need More Experts: Dynamic Routing in MoE Models
par: Huang, Quzhe, et autres
Publié: (2024)

Triplets Better Than Pairs: Towards Stable and Effective Self-Play Fine-Tuning for LLMs
par: Wang, Yibo, et autres
Publié: (2026)

DeepCode AI Fix: Fixing Security Vulnerabilities with Large Language Models
par: Berabi, Berkay, et autres
Publié: (2024)

AgentFold: Long-Horizon Web Agents with Proactive Context Management
par: Ye, Rui, et autres
Publié: (2025)

Memento: Fine-tuning LLM Agents without Fine-tuning LLMs
par: Zhou, Huichi, et autres
Publié: (2025)

MiCA Learns More Knowledge Than LoRA and Full Fine-Tuning
par: Rüdiger, Sten, et autres
Publié: (2026)

CoDAR: Continuous Diffusion Language Models are More Powerful Than You Think
par: Shen, Junzhe, et autres
Publié: (2026)

Graphene: Infrastructure Security Posture Analysis with AI-generated Attack Graphs
par: Jin, Xin, et autres
Publié: (2023)

Why LLMs Cannot Think and How to Fix It
par: Jahrens, Marius, et autres
Publié: (2025)

Less is More: Accurate Speech Recognition & Translation without Web-Scale Data
par: Puvvada, Krishna C., et autres
Publié: (2024)

SecureVibeBench: Benchmarking Secure Vibe Coding of AI Agents via Reconstructing Vulnerability-Introducing Scenarios
par: Chen, Junkai, et autres
Publié: (2025)

Can Performant LLMs Be Ethical? Quantifying the Impact of Web Crawling Opt-Outs
par: Fan, Dongyang, et autres
Publié: (2025)

Vulnerability of LLMs' Stated Beliefs? LLMs Belief Resistance Check Through Strategic Persuasive Conversation Interventions
par: Huang, Fan, et autres
Publié: (2026)

AgentBench: Evaluating LLMs as Agents
par: Liu, Xiao, et autres
Publié: (2023)

LLMs Struggle with Abstract Meaning Comprehension More Than Expected
par: Alhazmi, Hamoud, et autres
Publié: (2026)

Pruning Unsafe Tickets: A Resource-Efficient Framework for Safer and More Robust LLMs
par: Si, Wai Man, et autres
Publié: (2026)

A Survey on Retrieval-Augmented Text Generation for Large Language Models
par: Huang, Yizheng, et autres
Publié: (2024)

Long Is More Important Than Difficult for Training Reasoning Models
par: Shen, Si, et autres
Publié: (2025)