Research

Paper

AI LLM March 19, 2026

ClawTrap: A MITM-Based Red-Teaming Framework for Real-World OpenClaw Security Evaluation

Authors

Haochen Zhao, Shaoyang Cui

Abstract

Autonomous web agents such as \textbf{OpenClaw} are rapidly moving into high-impact real-world workflows, but their security robustness under live network threats remains insufficiently evaluated. Existing benchmarks mainly focus on static sandbox settings and content-level prompt attacks, which leaves a practical gap for network-layer security testing. In this paper, we present \textbf{ClawTrap}, a \textbf{MITM-based red-teaming framework for real-world OpenClaw security evaluation}. ClawTrap supports diverse and customizable attack forms, including \textit{Static HTML Replacement}, \textit{Iframe Popup Injection}, and \textit{Dynamic Content Modification}, and provides a reproducible pipeline for rule-driven interception, transformation, and auditing. This design lays the foundation for future research to construct richer, customizable MITM attacks and to perform systematic security testing across agent frameworks and model backbones. Our empirical study shows clear model stratification: weaker models are more likely to trust tampered observations and produce unsafe outputs, while stronger models demonstrate better anomaly attribution and safer fallback strategies. These findings indicate that reliable OpenClaw security evaluation should explicitly incorporate dynamic real-world MITM conditions rather than relying only on static sandbox protocols.

Metadata

arXiv ID: 2603.18762
Provider: ARXIV
Primary Category: cs.CR
Published: 2026-03-19
Fetched: 2026-03-20 06:02

Related papers

Raw Data (Debug)
{
  "raw_xml": "<entry>\n    <id>http://arxiv.org/abs/2603.18762v1</id>\n    <title>ClawTrap: A MITM-Based Red-Teaming Framework for Real-World OpenClaw Security Evaluation</title>\n    <updated>2026-03-19T11:14:45Z</updated>\n    <link href='https://arxiv.org/abs/2603.18762v1' rel='alternate' type='text/html'/>\n    <link href='https://arxiv.org/pdf/2603.18762v1' rel='related' title='pdf' type='application/pdf'/>\n    <summary>Autonomous web agents such as \\textbf{OpenClaw} are rapidly moving into high-impact real-world workflows, but their security robustness under live network threats remains insufficiently evaluated. Existing benchmarks mainly focus on static sandbox settings and content-level prompt attacks, which leaves a practical gap for network-layer security testing. In this paper, we present \\textbf{ClawTrap}, a \\textbf{MITM-based red-teaming framework for real-world OpenClaw security evaluation}. ClawTrap supports diverse and customizable attack forms, including \\textit{Static HTML Replacement}, \\textit{Iframe Popup Injection}, and \\textit{Dynamic Content Modification}, and provides a reproducible pipeline for rule-driven interception, transformation, and auditing. This design lays the foundation for future research to construct richer, customizable MITM attacks and to perform systematic security testing across agent frameworks and model backbones. Our empirical study shows clear model stratification: weaker models are more likely to trust tampered observations and produce unsafe outputs, while stronger models demonstrate better anomaly attribution and safer fallback strategies. These findings indicate that reliable OpenClaw security evaluation should explicitly incorporate dynamic real-world MITM conditions rather than relying only on static sandbox protocols.</summary>\n    <category scheme='http://arxiv.org/schemas/atom' term='cs.CR'/>\n    <category scheme='http://arxiv.org/schemas/atom' term='cs.AI'/>\n    <published>2026-03-19T11:14:45Z</published>\n    <arxiv:comment>8 pages, 5 figures, 2 tables. Preliminary technical report; quantitative experiments and extended evaluation to appear in v2</arxiv:comment>\n    <arxiv:primary_category term='cs.CR'/>\n    <author>\n      <name>Haochen Zhao</name>\n    </author>\n    <author>\n      <name>Shaoyang Cui</name>\n    </author>\n  </entry>"
}