Rushil Kubavat, home
← All workINDEX · 2025

Cyber Range Simulation

Two attackers vs three defenders, learning on a six-layer network.

RLQ-learningDockerLLM + RAG

A multi-agent reinforcement learning cyber range: two attacker agents and three defender agents learn against each other on a six-layer defense-in-depth network inside an isolated Docker lab. LLM + RAG guidance steers exploration.

Approach

  1. Multi-agent Q-learning: 2 attackers vs 3 defenders.
  2. Six-layer defense-in-depth topology.
  3. Fully isolated Docker lab environment.
  4. LLM + RAG guided exploration to escape naive random search.