Reproduce “The Critical Node Game” on OpenResearch

About this paper

Gabriele Dragotto, Amine Boukhtouta, Andrea Lodi, Mehdi Taobane

This research introduces the Critical Node Game (CNG), a game-theoretic model that quantifies cyber-security risk in cloud networks by framing attacker-defender interactions as a simultaneous two-player game. The model, along with a tailored algorithm, computes optimal defense strategies and assesses their efficiency and the attacker's degradation through custom metrics, successfully applying to real-world cloud network data.

Key insights

• The Critical Node Game models cyber-security as a simultaneous, non-cooperative game between an attacker and a defender, explicitly incorporating network topology and resource constraints.
• The Price of Security (PoS) and Price of Aggression (PoA) metrics quantify the efficiency of equilibrium strategies, showing the defender's payoff degradation due to stability and the attacker's payoff degradation, respectively.
• The 'ZERO Regrets' algorithm, tailored for this game, can compute high-quality approximate Nash Equilibria for large-scale networks, providing practically valuable security recommendations even when exact equilibria are computationally intensive.

Summary by alphaXiv.