TECH Signal 517
Show HN: Nightcrawler – A local AI pentesting agent running on a smartphone
For engineers building security tooling, Nightcrawler shows that full‑stack pentesting—recon, enumeration, exploitation, and report generation—can be performed without any cloud connectivity, reducing data‑exfiltration risk. Its on‑device AI decision loop and scope‑enforced safety layer illustrate a new approach to stealthy, low‑profile testing that avoids the noisy bursts typical of traditional scanners.
Written by elseif from the cluster below · every claim links back to a sourceThe three things worth knowing
The AI model (LFM2.5‑1.2B‑Instruct‑Heretic) runs locally on the phone’s Adreno 650 GPU via OpenCL, eliminating the need for external compute or internet access.
Nightcrawler follows a human‑like testing cadence, rotating across hosts with slow scan rates and cover traffic to stay under IDS/IPS detection thresholds.
The system requires a rooted Android device with Kali NetHunter, 12 GB+ RAM, and optionally a USB Wi‑Fi adapter for offline WPA2 cracking, and it provides a web dashboard for real‑time monitoring and control.
THE CLUSTER