AgentShield fingerprints AI agents via behavioral telemetry
Original title: AgentShield: Fortifying Web Applications Against AI Agents and Agentic Browsers
AgentShield detects autonomous agents through PRA-loop telemetry exposing differences from humans while mitigating Extinction-induced Policy Drift attacks.
LLM-powered browser agents autonomously interact with websites. Existing anti-bot defenses struggle against agentic systems. AgentShield uses dynamic web forms and interaction telemetry to capture artifacts in perception-reasoning-action loops. It introduces Extinction-induced Policy Drift where repeated obstructions relax agent safeguards. Defensive prompt injection exploits agent weaknesses for mitigation.
- Many defenses fail against AI-driven agents
- EPD causes agents to engage in unsafe behaviors
- Telemetry differentiates human from autonomous users
From the original article
1 min read at arxiv.orgLLM-powered browser agents are transforming web automation by autonomously interacting with websites and executing complex workflows. However, the effectiveness of existing anti-bot defenses against these systems remains unclear.
By Moiz Ali · arxiv.org
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.