Autopentest-drl Jun 2026

Multiple agents (red, green, blue) learning simultaneously in the same environment. Blue agents learn to patch, red agents learn to evade. This mirrors real cyber warfare and yields more robust defenses.

Download the source from the releases page and install dependencies: sudo -H pip install -r requirements.txt Use code with caution. Copied to clipboard autopentest-drl

You can run "Logical" mode to simply study attack paths on a virtual topology without firing real exploits, or "Real" mode to conduct actual penetration tests. Zero-Knowledge Start: Multiple agents (red

The Autopentest-DRL framework works as follows: autopentest-drl

: The system integrates with Metasploit via an RPC API (pymetasploit3) to execute the proposed attacks on a target network to verify exploitability.