When can decision models help LLM-driven network discovery?
When Do Decision Models Help LLM-Driven Network Discovery?
AI summary
Decision-model gating raised the local model's port recall on small networks from 75.5% to 89.0% and blocked decoy tarpits, but it slows early host discovery for frontier models. The study evaluated three network scales, n=10, 100, and 1000, in simulations calibrated on a 30-node containerized testbed: at large scale, chunked batch dispatch recovered 92% to 100% of open ports, while single-host dispatch dropped below 6% due to context saturation and scanning latency.