OpenClaw Defense Framework: 47 Adversarial Scenarios, 17% Native Defense Rate, HITL Layer Reaches 92%

arXiv·high signal

arXiv:2603.10387 systematically tests OpenClaw across six MITRE ATLAS/ATT&CK attack categories and finds native LLM-based defenses fail at an 83% rate—sandbox escape is the dominant vector. A Human-in-the-Loop (HITL) defense layer intercepts 8 severe attacks that bypassed all native defenses, lifting the overall defense rate to 19-92% depending on configuration. This is the most empirically grounded OpenClaw hardening guide published to date.