Engineering Secure Self-Adaptive Systems with Bayesian Games

Abstract Security attacks present unique challenges to self-adaptive system design due the adversarial nature of environment. Game theory approaches have been explored in security model malicious behaviors and reliable defense for a mathematically grounded manner. However, modeling as single player, done prior works, is insufficient under partial compromise fine-grained defensive strategies where rest with autonomy can cooperate mitigate impact attacks. To deal such issues, we propose new framework incorporating Bayesian game defender (i.e., system) at granularity components . Under attacks, architecture translated into multi-player , each component explicitly modeled an independent player while are encoded variant types components. The optimal strategy dynamically computed by solving pure equilibrium adaptation response) achieve best possible utility, improving resiliency against We illustrate our approach using example involving load balancing case study on inter-domain routing.