Training…
drag to pan · wheel/pinch to zoom

AI Attack Path Planner 2D: Q-Learning on a Network Graph

A reinforcement-learning agent explores a 2D graph of network hosts — from an internet-facing DMZ entry point through internal segments to a crown-jewel target — learning, hop by hop, which lateral-movement path best balances effort against the chance of tripping a defender's detection. Every edge updates its Q-value live via the Bellman equation, glowing brighter and thicker as the agent converges on a policy. Drag the canvas to pan the graph and scroll/pinch to zoom, while a reward sparkline and a per-edge Q-value bar chart along the bottom make the convergence measurable, not just visible. A defender-vigilance slider lets you watch the learned policy shift in real time as detection gets more expensive.