Anomaly detection using visualization and machine learning

Unauthorized access from inside or outside an organization has become a social problem in the last few years, making a system that can detect such accesses desirable. We therefore monitor normal activities using inductive logic programming (ILP) which is one of machine learning and detect anomalies. To ensure effective monitoring, we think the following two points must be considered. One point is automation of detection by ILP system, which is a rule generation engine, that always induces and updates effective rules. The other point is providing a visualization tool that reflects induced rules to the detection system. This tool enables an administrator to understand detection situations. For automated detection, we provide the ILP system with an automatic parameter adjustment function. For the visualization tool, we apply the visualization technology of a hyperbolic tree.