US Patent 11868478 System and method utilizing machine learning to predict security misconfigurations

Configuration monitoring is performed using a computer-based system and method by identifying misconfigured settings through the collection of large amounts of configuration data from diverse sources. The configuration data is then analyzed to identify misconfigured items. Automation of such configurations is implemented using machine learning to analyze existing configurations as well as new configurations. By using machine learning, the computer-based system and method can predict a pass state or a fail state of the configuration of a newly connected system in an organization. A logistic regression classifier is trained using old complying configuration data and data reflecting industry standards. The trained classifier can predict and classify whether a new configuration passes or fails the industry standards based on the training data of old configuration data. The computer-based system and method using the trained classifier generate and output a report indicating a pass or fail state of the new configuration.