Research
MUZZLE: Adaptive Agentic Red-Teaming of Web Agents Against Indirect Prompt Injection Attacks
Publication date not yet evaluated · Added arxiv.org
Page published
Coverage timeline
Single-source research — one report is available.
Why it matters
MUZZLE demonstrates that web agents remain systematically vulnerable to adaptive indirect prompt injection, giving defenders a method to automatically surface confidentiality, integrity, and availability violations that fixed-template evaluations miss.
MUZZLE is an automated agentic red-teaming framework, presented in a USENIX Security 2026 paper by researchers from Northeastern University and Mozilla, for evaluating the security of LLM-based web agents against indirect prompt injection attacks embedded in untrusted web content. Muzzle uses agent trajectories to identify high-salience injection surfaces and adaptively generates context-aware malicious instructions, discovering 44 new attacks across 4 web applications, including 3 cross-application prompt injection attacks and an agent-tailored phishing scenario.