Reinforcement Learning for Generating Secure Configurations

Many security problems in software systems are because of vulnerabilities caused by improper configurations. A poorly configured system leads to a multitude that can be exploited adversaries. The problem becomes even more serious when the architecture underlying is static and misconfiguration remains for longer period time, enabling adversaries thoroughly inspect under attack during reconnaissance stage. Employing diversification techniques such as Moving Target Defense (MTD) minimize risk exposing vulnerabilities. MTD an evolving defense technique through which surface continuously changing. However, effectiveness dynamically changing platform depends not only on goodness next configuration setting with respect minimization surfaces but also diversity set configurations generated. To address generating diverse large secure configurations, this paper introduces approach based Reinforcement Learning (RL) agent trained generate desirable reports performance RL-based some case studies.