ClawTrap: A MITM-Based Red-Teaming Framework for Real-World OpenClaw Security Evaluation
This addresses a practical gap in network-layer security testing for high-impact real-world workflows, though it is incremental as it builds on existing red-teaming concepts.
The paper tackles the problem of insufficient security evaluation for autonomous web agents like OpenClaw under live network threats by introducing ClawTrap, a MITM-based red-teaming framework, and finds that weaker models are more vulnerable to tampered observations while stronger models show better anomaly handling.
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.