Manipulating LLM Web Agents with Indirect Prompt Injection Attack via HTML Accessibility Tree
By: Sam Johnson, Viet Pham, Thai Le
Potential Business Impact:
Hackers can trick web robots into doing bad things.
This work demonstrates that LLM-based web navigation agents offer powerful automation capabilities but are vulnerable to Indirect Prompt Injection (IPI) attacks. We show that adversaries can embed universal adversarial triggers in webpage HTML to hijack agent behavior that utilizes the accessibility tree to parse HTML, causing unintended or malicious actions. Using the Greedy Coordinate Gradient (GCG) algorithm and a Browser Gym agent powered by Llama-3.1, our system demonstrates high success rates across real websites in both targeted and general attacks, including login credential exfiltration and forced ad clicks. Our empirical results highlight critical security risks and the need for stronger defenses as LLM-driven autonomous web agents become more widely adopted. The system software (https://github.com/sej2020/manipulating-web-agents) is released under the MIT License, with an accompanying publicly available demo website (http://lethaiq.github.io/attack-web-llm-agent).
Similar Papers
In-Browser LLM-Guided Fuzzing for Real-Time Prompt Injection Testing in Agentic AI Browsers
Cryptography and Security
Finds hidden tricks that trick web robots.
Adaptive Attacks Break Defenses Against Indirect Prompt Injection Attacks on LLM Agents
Cryptography and Security
Breaks AI security, letting hackers control it.
Realistic Environmental Injection Attacks on GUI Agents
Cryptography and Security
Makes websites trick computer helpers into making mistakes.