Fecha de publicación: 28/01/2021 Fuente: WIPO "hive" Systems and methods for network security testing of target computer networks using AI neural networks. A command and control server controls a number of geographically separated processors running a number of neural networks. A central data hive is accessible to all the processors. The processors are organizable into logical hemisphere groupings for specific tasks and/or projects. For security testing, hemisphere groupings are created for the project. Based on data for the target system on the data hive, attacks are formulated by a hemisphere grouping and these potential attacks are tested against known characteristics of the target network. Validated potential attacks and, in some cases, random attacks, are executed and data generated by the executed attacks are stored in the data hive for use in formulating and executing other further attacks. Potential attacks may involve mining social media networks for data on users of the target system.