Skip to content
Isaac Thor edited this page Mar 8, 2026 · 2 revisions

RMLW Wiki

Purpose

The Recursive Meta-Learning Workbench (RMLW) is a security research tool for web vulnerability testing in authorized lab environments. It combines a static Web Attack Workbench (XSS, SQLi, LFI, CMDi, SSRF) with a recursive meta-learning layer that adapts payloads and test strategies based on target feedback.

High-Level Architecture

RMLW has two main layers:

  1. Web Attack Workbench – Runs vulnerability tests against configured endpoints. Tests include XSS (reflection), SQLi (boolean logic), LFI (file inclusion), CMDi (command injection), and SSRF (internal access probes).

  2. Recursive Meta-Learning Layer – Orchestrates the workbench using:

    • EnvironmentModel: Captures target characteristics (WAF, filters, successful payloads)
    • RLPathOptimizer: Q-learning to choose test families
    • EvolutionaryPayloadGenerator: Mutates and evolves payloads
    • PersistentArchive: Stores successful payloads for reuse
    • MetaController: UCB bandit selects which mechanism to apply

The MetaController selects a mechanism (RL, evolution, archive, or baseline), the workbench runs tests, findings update the model and archive, and the cycle repeats.

Wiki Pages

  • Getting-Started – Prerequisites, install, baseline scan, learning mode, troubleshooting
  • Architecture – Detailed component overview and data flow
  • Contributing – How to run tests, coding style, PR process
  • Security-and-Scope – Authorized use, scope, assumptions

Clone this wiki locally