Adversarial Images Hijack Web Agents from Visual Grounding to Browser Execution
Wanjing Han , Levi Taiji Li , Mu Zhang , Yue Jiang , Guanhong Tao
- 🏛 Institutions
- University of Utah
- 📅 Date
- October 7, 2026
- 📑 Publisher
- arXiv
- 💻 Env
- Web
- 🔑 Keywords
TLDR
Studies end-to-end visual adversarial attacks targeting multimodal web agents from visual grounding down to browser execution. Demonstrates that imperceptible adversarial image perturbations on webpages consistently hijack agent attention and cause targeted misclicks across Mind2Web and VisualWebArena.
Related papers (24)
- When Charts Deceive Web Agents: A Paired Benchmark for Misleading Visualization-Grounded TasksOctober 9, 2026 · arXiv
- InterA11y: Interactive Web Accessibility Error Detection with LLM AgentsOctober 29, 2026 · arXiv
- AgentHeal: Evaluating Goal-Preserving Resource Recovery of Web Agents under Access BarriersOctober 29, 2026 · arXiv
- WebAnything: From the Wild Web to Scalable Agent EnvironmentsOctober 27, 2026 · arXiv
- SkillLifeBench: Benchmarking Security Across the Lifecycle of Agent SkillsOctober 26, 2026 · arXiv
- Back to Basics: A Skill-Centric Approach to Generalist Web AgentsOctober 26, 2026 · arXiv
- AgentProofBench: Benchmarking Reusable Verifiers for Web AgentsOctober 25, 2026 · arXiv
- SkillShift-Bench: Benchmarking Skill Learning under Environment Evolution for Web AgentsOctober 24, 2026 · arXiv
- Learning What to Ask for Tomorrow: Active Preference Acquisition for Long-Term Personalized Web AgentsOctober 22, 2026 · arXiv
- Step-Aligned Script Generation with Verification and Recovery for Web Automation TestingOctober 19, 2026 · arXiv
- LoWAgent: A Low-Resource Multilingual Benchmark for Web AgentsOctober 19, 2026 · arXiv
- Why Generate a Click? Plug-in Decision Heads for Web AgentsOctober 14, 2026 · arXiv
- Teaching Agents in their Own Style: Blackbox Teacher Distillation for Web AgentsOctober 14, 2026 · arXiv
- STP-GUI: Spatially-Aware Token Pruning for Efficient Web-GUI AgentsOctober 14, 2026 · arXiv
- MHVF: Training-Free GUI Grounding via Verification-Guided Multi-Hypothesis SearchOctober 13, 2026 · arXiv
- Seg2Click: Learning Pixel-Level Action Masks for GUI Visual GroundingOctober 12, 2026 · arXiv
- Web$hell: A Unified Interface for Agentic Web AccessOctober 9, 2026 · arXiv
- AdvSim2Real : Training Web Agents Against Adaptive Prompt Injection in a Web World ModelOctober 6, 2026 · arXiv
- DashAct: A Progressive Diagnostic Benchmark for GUI Agents in Interactive Dashboard AnalysisSeptember 26, 2026 · arXiv
- WebChallenger: A Reliable and Efficient Generalist Web AgentJune 9, 2026 · arXiv
- How do Visual Attributes Influence Web Agents? A Comprehensive Evaluation of User Interface Design FactorsJanuary 29, 2026 · arXiv
- Web Agent Agentic Reinforcement Learning Decision Model Under Multi-Cost and Failure RiskJanuary 12, 2026 · arXiv
- Genesis: Evolving Attack Strategies for LLM Web Agent Red-TeamingOctober 21, 2025 · ICME 2026
- Iris: Breaking GUI Complexity with Adaptive Focus and Self-RefiningDecember 13, 2024 · arXiv