Training a pentesting agent from scratch is notoriously brittle. The reward signal is extremely sparse – an agent might flail for 5,000 episodes with zero reward before accidentally discovering a vulnerability. Researchers solve this via .
: It uses Nmap to scan networks and determine existing vulnerabilities in real-time. autopentest-drl
We trained AutoPentest-DRL on a simulated corporate network (30 hosts, 4 subnets) for 50,000 episodes. Training a pentesting agent from scratch is notoriously